AWS Solutions Architect Practice Test

AWS Certified Solutions Architect – Associate

AWS Solutions Architect overview

  • 200 practice questions
  • 32 flashcards
  • 4 study guide topics
  • 37 key terms

What you'll study

  • Design Secure Architectures25%
  • Design Resilient Architectures25%
  • Design High-Performing Architectures25%
  • Design Cost-Optimized Architectures25%

The AWS Certified Solutions Architect – Associate is a foundational-to-intermediate AWS certification aimed at people who design systems on the AWS Cloud. It suits cloud engineers, developers and system administrators stepping into architecture work, as well as technical consultants and pre-sales staff who need to translate requirements into working designs. The exam is organised around four equally weighted design domains: secure architectures, resilient architectures, high-performing architectures, and cost-optimized architectures. Questions are presented in multiple-choice and multiple-response formats, and each scenario asks you to pick the design that best satisfies a stated set of requirements, often with an explicit constraint such as budget, latency, or compliance. Preparation works best as a mix of reading and doing: study the AWS Well-Architected Framework and the core service documentation for compute, storage, networking, databases, identity and monitoring, then build and break small environments so that trade-offs become intuitive. Practise reading the question stem for the decisive constraint before scanning the answer options, and rehearse eliminating answers that are technically valid but over-engineered or off-requirement. This page is independent study material and is not affiliated with, endorsed by, or sponsored by Amazon Web Services; AWS and related marks belong to their owner. Always confirm exam logistics on the official AWS exam guide before booking.

Common mistakes

When a question says an application running on EC2, Lambda, or ECS needs to call S3, DynamoDB, or SQS, the candidate picks 'create an IAM user, generate an access key, store it in the app config / Secrets Manager' instead of an IAM role.

Access-key habits come from on-premises credential management, and the phrase 'programmatic access' in an option triggers the IAM-user answer even when the workload runs inside AWS and a role is available.

What to do instead

Before evaluating options, ask one question: 'Where does the code run?' Inside AWS compute = IAM role (instance profile / task role / execution role), no stored credentials. Outside AWS = STS AssumeRole, IAM Roles Anywhere, or federation. Static access keys are almost never the exam answer except for third-party tooling you cannot change.

The candidate uses a security group to explicitly deny traffic from a single bad IP, or uses a network ACL to control traffic between two instances in the same subnet, and then forgets to add the ephemeral-port return rule for a NACL.

The rule 'security groups are allow-only, NACLs allow and deny' is memorized as a slogan but not applied to the two real axes: scope (ENI/instance vs subnet) and statefulness (stateful vs stateless, so NACLs need both directions on ephemeral ports 1024-65535).

What to do instead

Run a two-question filter on every network question: (1) Is the control point an ENI/instance or a whole subnet? Instance-level = security group; subnet-level or 'block a specific IP range' = NACL. (2) Is it stateless? If NACL, write both inbound and outbound rules including the ephemeral port range. Also remember SGs can reference other SGs, which is how you allow instance-to-instance without CIDRs.

The stem says 'improve read performance for a read-heavy database' and the candidate picks Multi-AZ, or the stem says 'survive an Availability Zone failure with zero data loss' and the candidate picks a Read Replica.

RDS Multi-AZ and Read Replicas are both described as 'a second copy of the database', so the features collapse into one memory. The candidate pattern-matches the words 'high availability' or 'second copy' without separating synchronous standby from asynchronous readable copy.

What to do instead

Memorize a three-row table and map the stem's keyword to a row before looking at options. Multi-AZ: synchronous, standby not readable, automatic failover, purpose = availability. Read Replica: asynchronous, readable, manual promotion, purpose = read scaling. Multi-Region / cross-region replica: purpose = regional DR. Also note Aurora has its own rules (Aurora Replicas can fail over, Global Database for cross-region).

The candidate treats a service as automatically covering every failure scope, e.g. 'S3 is eleven nines durable, so no DR plan is needed' or 'Lambda is serverless, so it is multi-region', or assumes a single DynamoDB table is already global.

Durability, availability, and geographic blast radius get conflated. Marketing numbers about durability are remembered, but the region/AZ boundary of each service is not.

What to do instead

For any resilience question, first label the blast radius the stem requires (instance, AZ, region, global) and then label each candidate service's scope: S3 bucket = regional (cross-region replication is a separate feature), EBS volume = single AZ, DynamoDB table = regional (global tables for multi-region), Lambda = regional. Eliminate any option whose native scope is smaller than the requirement unless it is paired with the feature that extends it.

The candidate answers a decoupling question with SQS where fan-out to multiple consumers is required, or with SNS where durable buffering, retry, and a backlog for a slow consumer are required, or ignores the need for a dead-letter queue.

SQS and SNS are both 'messaging services', so the choice default is whatever service was studied most recently. The distinguishing properties (pull vs push, single consumer group vs many subscribers, retention, replay) are not recalled under time pressure.

What to do instead

Memorize the shape of each: SQS = queue, pull, one consumer group competing for messages, retention up to 14 days, needs a DLQ for poison messages. SNS = pub/sub push, fan-out to many subscribers at once, no retention for failed delivery unless you add a queue. EventBridge = routing and filtering by event pattern, plus SaaS sources. SQS FIFO = ordering and deduplication. If the stem says 'send the same event to five systems', the answer is fan-out, not a single queue.

The stem shows a performance bottleneck and the candidate's answer is 'move to a larger instance type', even when read replicas, ElastiCache, CloudFront, or a queue are among the options.

Scaling up is the intuitive and most familiar answer from on-premises experience, and the candidate does not spend time deciding what the bottleneck actually is (CPU, read I/O, write I/O, connection count, or client latency).

What to do instead

Classify the bottleneck before comparing options: CPU-bound = bigger instance or horizontal scaling; read-heavy database = read replicas or a cache in front; repeated identical reads / static content = CloudFront or ElastiCache; write-heavy = queue writes, partition key/sharding, or DynamoDB on-demand; long-running synchronous calls between services = decouple with SQS. Pick the option that targets the labeled bottleneck, not the one that adds raw capacity everywhere.

For a global performance or failover requirement, the candidate picks Route 53 latency-based routing for a non-HTTP TCP workload, or picks CloudFront when the requirement is static anycast IP addresses for allowlisting, or assumes CloudFront provides fixed IPs.

CloudFront, Global Accelerator, and Route 53 routing policies are all 'global and fast', so they collapse into one memory. The differentiators (HTTP(S) content caching vs TCP/UDP acceleration with anycast IPs vs DNS-level resolution) are not separated.

What to do instead

Use a decision tree: cacheable or HTTP(S) content delivered to worldwide users = CloudFront; non-HTTP(S) protocols, static anycast IPs, or fast regional failover without DNS TTL delay = Global Accelerator; multi-region routing based on latency, geography, or weights with DNS only = Route 53 policies. Record that CloudFront does not give you a fixed IP list (use a managed prefix list or Global Accelerator if IP allowlisting is required).

The candidate chooses a storage service by capacity or price rather than by access semantics, e.g. EBS for a shared directory mounted on several Linux instances, EFS for a single-instance high-IOPS database, or S3 for a low-latency POSIX filesystem.

Block, file, and object storage are all 'storage' in the candidate's mental model, and EFS sounds like a newer, better EBS. The key signals in the stem (shared, POSIX, SMB, single-AZ, random I/O, HPC scratch) are not mapped to services.

What to do instead

Anchor on access semantics: block storage attached to one instance in one AZ with high random IOPS = EBS (gp3/io2); shared POSIX NFS across Linux instances and AZs = EFS; native Windows/SMB or Active Directory integration = FSx for Windows; HPC scratch with sub-millisecond latency = FSx for Lustre; object with HTTP access and lifecycle tiering = S3; ephemeral ultra-high-IOPS scratch = instance store. Look for the words 'shared', 'POSIX', 'SMB', and 'single AZ' in the stem as direct pointers.

The candidate picks a lower storage-class price per GB (S3 One Zone-IA, Glacier Instant Retrieval, or Deep Archive) for data that is accessed frequently, or compares NAT Gateway, cross-AZ transfer, or provisioned capacity using only the headline hourly rate.

Price-per-GB is the number candidates remember from study material, while retrieval fees, minimum storage durations, request charges, per-GB processing, and cross-AZ data transfer are the numbers that decide the real bill.

What to do instead

Compute a total cost checklist before choosing: storage rate x GB, plus request count charges, plus retrieval fees per GB, plus minimum billable duration (30 days for Standard-IA/One Zone-IA, 90 days for Glacier Instant Retrieval/Flexible Retrieval, 180 days for Deep Archive), plus data transfer in/out and cross-AZ. If the data is accessed more than roughly once a month or is smaller than 128 KB per object, Standard usually wins. For NAT Gateway, price per hour plus per-GB processed, and remember it is per AZ.

The candidate answers 'On-Demand for everything' to avoid commitment risk, or proposes Spot Instances for a stateful primary database or a workload that cannot tolerate interruption, or ignores Savings Plans and Reserved capacity entirely.

There is a bias toward the safest-sounding option, and commitment-based pricing feels risky, so the candidate eliminates the correct cost answer. Conversely, 'cheapest compute' is chosen when the words Spot appear, without checking fault tolerance.

What to do instead

Split the workload into three buckets and price each: steady-state baseline that runs 24/7 = Compute Savings Plans or Reserved Instances (and DynamoDB reserved capacity, ElastiCache reserved nodes); unpredictable or short-lived = On-Demand; fault-tolerant, checkpointed, interruption-tolerant batch/CI = Spot. If the stem gives a stable utilization percentage, that is a direct signal for a commitment-based option. If the stem says 'stateful', 'cannot be interrupted', or 'single database', eliminate Spot.

Exam requirements can change. Confirm current details with the exam provider.

AWS Solutions Architect practice

200 questions. Each one lists the choices, the correct answer, and an explanation. With JavaScript, this list opens one question at a time. The checker at the top of this page saves progress in this browser when it is available.

  1. Question 1Design Secure Architectures

    An IAM user belongs to a group whose policy allows s3:GetObject on a bucket. A separate inline user policy explicitly denies s3:GetObject on the same bucket. What is the outcome?

    1. Allowed, but only when MFA is present.
    2. Denied, because explicit deny wins. Correct answer.
    3. Allowed, because the group policy grants it.
    4. Denied only if the bucket policy also denies it.

    Explanation

    IAM evaluates all applicable policies and any explicit deny anywhere in the chain overrides every allow. Group membership does not change that precedence, and a bucket policy deny is not required for the inline deny to take effect.

  2. Question 2Design Secure Architectures

    Which measure most directly reduces the risk of compromise of the AWS account root user?

    1. Attach an administrator policy to the root user.
    2. Delete the root access key and use it only for CLI work.
    3. Enable MFA and avoid creating root access keys. Correct answer.
    4. Store root credentials in Secrets Manager for rotation.

    Explanation

    The root user cannot be restricted by IAM policies, so the practical controls are strong MFA on the account and not creating long-lived root access keys. Routine administrative work should be performed with IAM roles or users instead.

  3. Question 3Design Secure Architectures

    A web tier must accept HTTPS from the internet, and a database tier must accept traffic only from the web tier. Which configuration enforces this with security groups?

    1. Place both tiers in one group open to the VPC CIDR.
    2. Allow 0.0.0.0/0 on the database group and filter at the OS.
    3. Allow the web tier's private IP range in the database group.
    4. Use the web tier's security group as the source. Correct answer.

    Explanation

    A security group rule can name another security group as its source, so instances are permitted based on group membership rather than IP addresses. That keeps the rule valid as the web tier scales or changes addresses.

  4. Question 4Design Secure Architectures

    Which Amazon S3 default encryption option produces per-object key usage entries in CloudTrail and lets you govern access with a key policy?

    1. Client-side encryption
    2. SSE-S3
    3. SSE-C
    4. SSE-KMS Correct answer.

    Explanation

    SSE-KMS uses AWS KMS keys whose usage is logged in CloudTrail and whose access is governed by key policies and IAM. SSE-S3 and SSE-C use keys that are not customer-governed KMS keys with per-key audit trails.

  5. Question 5Design Secure Architectures

    An application running on Amazon EC2 must read objects from Amazon S3. What is the secure way to provide credentials?

    1. Create an IAM user and embed keys in the code.
    2. Put the root access keys in user data.
    3. Ship long-lived keys inside a shared encrypted AMI.
    4. Attach an IAM role to the instance profile. Correct answer.

    Explanation

    An instance profile delivers temporary, automatically rotated credentials to the instance through the instance metadata service. Embedding static keys in code, user data, or AMIs spreads long-lived secrets that are hard to revoke.

  6. Question 6Design Secure Architectures

    Which AWS service records API activity in an account for security auditing and incident investigation?

    1. AWS CloudTrail Correct answer.
    2. Amazon CloudWatch
    3. AWS Trusted Advisor
    4. AWS Config

    Explanation

    CloudTrail logs API calls, including the identity, source IP, and time of each request, which supports auditing and forensics. CloudWatch handles metrics and logs, Config tracks resource configuration, and Trusted Advisor gives best-practice checks.

  7. Question 7Design Secure Architectures

    An S3 bucket with versioning enabled must be protected against permanent deletion of object versions. Which feature adds that protection?

    1. Enable default encryption with SSE-S3.
    2. Enable MFA Delete on the bucket. Correct answer.
    3. Enable S3 Transfer Acceleration for uploads.
    4. Require presigned URLs for every delete call.

    Explanation

    MFA Delete requires an additional authentication factor before object versions or the bucket versioning state can be permanently removed. Versioning alone only keeps prior versions, it does not stop an authorized caller from deleting them.

  8. Question 8Design Secure Architectures

    Which service is purpose-built to store and automatically rotate database credentials?

    1. AWS Secrets Manager Correct answer.
    2. S3 Object Lock
    3. AWS Config
    4. IAM Access Analyzer

    Explanation

    Secrets Manager stores secrets and can rotate them on a schedule using a Lambda rotation function, with native support for Amazon RDS and Amazon Redshift credentials. The other services address compliance, access analysis, and immutability.

  9. Question 9Design Secure Architectures

    An IAM user has an identity policy granting s3:* and a permissions boundary that allows only s3:GetObject. What are the user's effective S3 permissions?

    1. No S3 access at all
    2. All s3:* actions except s3:DeleteObject
    3. All s3:* actions from the identity policy
    4. Only s3:GetObject Correct answer.

    Explanation

    Effective permissions are the intersection of identity-based policies and the permissions boundary, so only the actions allowed by both remain. Here the boundary limits the user to s3:GetObject.

  10. Question 10Design Secure Architectures

    An organization has an SCP on an OU that denies ec2:RunInstances for all member accounts, and a user in one member account has an administrator policy. Can that user launch an EC2 instance?

    1. Yes, an administrator policy always wins inside the account.
    2. Yes, if the user has MFA configured.
    3. Yes, if the user signs in as the account root user.
    4. No, because SCPs cap account permissions. Correct answer.

    Explanation

    Service control policies set the maximum permissions available to principals in member accounts, and identity policies can only grant a subset of that ceiling. An explicit deny in an SCP therefore blocks the action regardless of in-account administrator access.

  11. Question 11Design Secure Architectures

    A team in account A must read objects from a bucket in account B. Which combination enables cross-account access?

    1. An IAM user in B whose keys are shared with A.
    2. A VPC peering connection between the two accounts.
    3. A bucket ACL that grants A's canonical user.
    4. A role in B trusting A, plus sts:AssumeRole in A. Correct answer.

    Explanation

    Cross-account access is normally granted by a role in the trusting account whose trust policy names the other account, plus permission in the trusted account to call sts:AssumeRole. The caller then receives temporary credentials scoped to that role.

  12. Question 12Design Secure Architectures

    Instances in a private subnet must reach Amazon S3 without using the public internet, and cost should be minimized. Which solution fits?

    1. An interface endpoint in each subnet.
    2. A gateway VPC endpoint with a route table entry. Correct answer.
    3. A site-to-site VPN to an S3 public endpoint.
    4. A NAT gateway in each Availability Zone.

    Explanation

    A gateway endpoint for S3 attaches to route tables, keeps traffic on the AWS network, and has no hourly or data processing charge. Interface endpoints work but create billable elastic network interfaces in each subnet.

  13. Question 13Design Secure Architectures

    A bucket policy allows public read access, but the account-level S3 Block Public Access setting is enabled. What is the result?

    1. The policy applies but only to IAM users.
    2. Only ACL-based public access is blocked.
    3. The public policy is blocked. Correct answer.
    4. The bucket policy overrides the block.

    Explanation

    Account-level Block Public Access overrides bucket policies and ACLs that would grant public access, so the public statement does not take effect. The block must be removed at the account or bucket level before public access is possible.

  14. Question 14Design Secure Architectures

    A partner needs temporary download access to a single S3 object for one hour, and no IAM users should be created. Which approach meets the requirement?

    1. Create an IAM user per partner and rotate keys.
    2. Make the object public, then revert after one hour.
    3. Generate a presigned URL with a short expiration. Correct answer.
    4. Serve the object from an unauthenticated CloudFront distribution.

    Explanation

    A presigned URL carries the signer's permissions for a limited time and can be scoped to one object, so no long-lived identity is needed. Making the object public exposes it to anyone who has the URL, with no authentication.

  15. Question 15Design Secure Architectures

    An IAM policy in account A grants kms:Decrypt on a customer managed key owned by account B, but the key policy does not mention account A. What happens when account A calls Decrypt?

    1. Decrypt succeeds after an IAM policy refresh delay.
    2. Denied only if the key is asymmetric.
    3. Decrypt succeeds.
    4. Denied, because the key policy must allow the account. Correct answer.

    Explanation

    AWS KMS requires that the key policy itself delegate access to the external account before IAM policies in that account can allow the action. Without that delegation, the decrypt call fails no matter what the caller's IAM policy says.

  16. Question 16Design Secure Architectures

    A Lambda function must write items to one DynamoDB table. Which permission design follows least privilege?

    1. Attach a role with only the needed DynamoDB actions. Correct answer.
    2. Attach AdministratorAccess to the function's role.
    3. Store IAM user keys in environment variables.
    4. Use the account root credentials inside the function.

    Explanation

    The execution role should grant only the specific DynamoDB actions on the specific table resource, which is the minimum required by the function. Broader permissions or static keys increase the blast radius of a compromise.

  17. Question 17Design Secure Architectures

    Which statement about security groups and network ACLs is accurate?

    1. Both are stateless and support allow and deny rules.
    2. Both are stateful and evaluated in numeric order.
    3. NACLs are stateful; security groups are stateless.
    4. Security groups are stateful; NACLs are stateless. Correct answer.

    Explanation

    Security groups are stateful, so return traffic for an allowed connection is automatically permitted. Network ACLs are stateless, so both inbound and outbound rules must explicitly allow the traffic.

  18. Question 18Design Secure Architectures

    Administrators need interactive shell access to EC2 instances in private subnets without a bastion host or an open SSH port. Which approach satisfies this?

    1. Put instances in a public subnet with an EIP.
    2. Use EC2 Instance Connect with a shared key.
    3. Publish a bastion host with a public elastic IP.
    4. Use SSM Session Manager with an instance profile. Correct answer.

    Explanation

    Session Manager reaches instances through the SSM agent and an instance profile, so no inbound ports, bastion, or public addresses are needed. Access is logged and controlled through IAM policies.

  19. Question 19Design Secure Architectures

    A mobile app needs user sign-in through social identity providers and must exchange those identities for scoped AWS credentials used by API Gateway. Which design fits?

    1. SAML federation with an external IdP for each user.
    2. AWS IAM Identity Center permission sets per user.
    3. Create one IAM user per application user.
    4. Cognito user pool with an identity pool. Correct answer.

    Explanation

    A Cognito user pool handles authentication with social providers and issues tokens, while an identity pool exchanges those tokens for temporary AWS credentials scoped to an IAM role. This avoids creating an IAM identity per application user.

  20. Question 20Design Secure Architectures

    A public web application must be protected against SQL injection attempts and large-scale DDoS attacks at the edge. Which combination addresses both?

    1. GuardDuty findings routed to Security Hub.
    2. AWS WAF on CloudFront plus Shield Advanced. Correct answer.
    3. NACLs with deny rules for malicious IP addresses.
    4. An S3 bucket policy that restricts referrers.

    Explanation

    AWS WAF provides managed and custom rules such as SQL injection match conditions, and Shield Advanced adds DDoS detection and response for edge distributions. Together they cover application-layer filtering and volumetric attack mitigation.

  21. Question 21Design Secure Architectures

    A SaaS provider assumes an IAM role in each customer's account. Which control prevents a third party from tricking the provider into using a role session meant for a different customer?

    1. iam:PassedToService in the session policy.
    2. aws:PrincipalOrgID in the bucket ACL.
    3. aws:SourceIp in the permission policy.
    4. sts:ExternalId in the trust policy condition. Correct answer.

    Explanation

    The external ID is a value shared between the provider and the specific customer, and requiring it in the role trust policy's condition blocks the confused deputy problem. A caller who does not know that customer's external ID cannot assume the role.

  22. Question 22Design Secure Architectures

    A federated workforce must be able to access only the resources tagged with the user's department, and departments change frequently. Which design scales best?

    1. SCPs per OU that match department names.
    2. One role plus a session tag with a tag-based condition. Correct answer.
    3. One IAM role per department, each with separate policies.
    4. A bucket policy listing every department ARN.

    Explanation

    Attribute-based access control maps identity attributes to session tags and uses conditions such as aws:PrincipalTag to compare them with resource tags. Adding a department then requires no new policy or role because the rule is expressed in terms of tags.

  23. Question 23Design Secure Architectures

    A company uses an S3 gateway VPC endpoint. Which statement correctly describes its scope?

    1. It applies only to route tables in its own VPC. Correct answer.
    2. It can be used from peered VPCs and on-premises networks.
    3. It requires a NAT gateway to function correctly.
    4. It supports only HTTPS on port 443 from EC2 instances.

    Explanation

    A gateway endpoint is associated with route tables in the VPC where it is created and cannot be reached from peered VPCs, VPN connections, or Direct Connect. Workloads outside that VPC need an interface endpoint or another path.

  24. Question 24Design Secure Architectures

    A service must be given temporary programmatic permission to use a customer managed KMS key without editing the key policy. Which mechanism should be used?

    1. Create a KMS grant on the key. Correct answer.
    2. Publish a new version of the key policy.
    3. Attach a resource-based policy to the calling service.
    4. Attach an IAM permissions boundary to the service role.

    Explanation

    A KMS grant delegates specific cryptographic operations on a key to a principal, optionally with constraints and an expiration, without modifying the key policy. Grants are the intended mechanism for temporary, programmatic delegation.

  25. Question 25Design Secure Architectures

    A security team must continuously identify S3 buckets and IAM roles in an account that are shared with principals outside the account. Which tool provides this?

    1. IAM Access Analyzer findings for the zone of trust. Correct answer.
    2. CloudTrail Insights event analysis.
    3. Trusted Advisor cost optimization checks.
    4. AWS Config rules aggregated with an aggregator account.

    Explanation

    IAM Access Analyzer evaluates resource-based policies against a zone of trust and reports resources reachable by principals outside that zone. Findings update as policies change, which supports continuous external-access monitoring.

  26. Question 26Design Secure Architectures

    A company wants to detect anomalous API calls, cryptocurrency mining activity, and communication with known malicious IP addresses across its AWS accounts with minimal configuration. Which service should it use?

    1. AWS Config
    2. Amazon GuardDuty Correct answer.
    3. AWS Trusted Advisor
    4. Amazon Inspector

    Explanation

    GuardDuty is a managed threat-detection service that continuously analyzes CloudTrail management and data events, VPC Flow Logs, and DNS logs for malicious or anomalous behavior. It requires no agents or infrastructure to deploy.

  27. Question 27Design Secure Architectures

    A request to read an object is allowed by an identity-based policy and by a bucket policy, but an SCP applied in the account contains a Deny statement for that action. What is the outcome?

    1. Allowed, provided the caller is an IAM role rather than an IAM user.
    2. Denied, because SCPs replace all identity-based policies in member accounts.
    3. Denied, because an explicit deny always overrides any allow. Correct answer.
    4. Allowed, because both the bucket policy and the identity policy allow the action.

    Explanation

    AWS evaluates all applicable policies together and the result is the union of allows, minus anything explicitly denied. An explicit Deny in any policy type, including an SCP, always wins over allow statements elsewhere.

  28. Question 28Design Secure Architectures

    Which statement describes AWS Shield Standard?

    1. It must be turned on per Region and is billed for each protected resource.
    2. It provides 24/7 access to the DDoS Response Team and cost protection against DDoS-related scaling charges on protected resources.
    3. It inspects HTTP requests for SQL injection and cross-site scripting patterns.
    4. It is on by default for all accounts at no cost and blocks common volumetric attacks. Correct answer.

    Explanation

    Shield Standard is automatically applied to all AWS customers at no additional charge and mitigates common network and transport layer DDoS attacks. Advanced capabilities such as the DDoS Response Team and cost protection require the paid Shield Advanced tier.

  29. Question 29Design Secure Architectures

    A subnet's network ACL contains rules numbered 100 (allow), 200 (deny), and the default asterisk rule. How is inbound traffic evaluated?

    1. Rules are processed in the order they were created, and the most recent rule wins.
    2. The most specific rule, based on the narrowest CIDR range, is applied first.
    3. All matching rules are combined, and any allow overrides a deny.
    4. Rules are processed in ascending rule-number order and the first match decides. Correct answer.

    Explanation

    Network ACLs are ordered, stateless filters: each rule is evaluated from the lowest rule number upward and the first rule that matches the traffic determines allow or deny. Rule 100 would therefore match before rule 200.

  30. Question 30Design Secure Architectures

    An application running on Amazon EC2 must call AWS APIs. What is the recommended way to supply credentials to the instance?

    1. Bake an IAM user's access keys into the AMI used to launch instances.
    2. Attach an IAM role to the instance through an instance profile. Correct answer.
    3. Store long-term access keys in a file on the instance's EBS volume.
    4. Pass access keys to the instance in the user data script at launch.

    Explanation

    An instance profile delivers IAM role credentials to the instance, and the AWS SDK retrieves and rotates those temporary credentials automatically. This avoids distributing long-term keys that cannot be scoped or expired easily.

  31. Question 31Design Secure Architectures

    An auditor asks for AWS's SOC 2 report and ISO certifications. Which service provides self-service access to these documents?

    1. AWS Security Hub
    2. AWS Artifact Correct answer.
    3. AWS Config
    4. AWS Audit Manager

    Explanation

    AWS Artifact is the self-service portal for downloading AWS compliance reports and agreements such as SOC, ISO, PCI, and HIPAA artifacts. Audit Manager instead automates evidence collection for your own audits.

  32. Question 32Design Secure Architectures

    A company needs to discover and classify personally identifiable information stored in Amazon S3 buckets. Which service is designed for this?

    1. AWS Shield Advanced
    2. Amazon Macie Correct answer.
    3. Amazon GuardDuty
    4. Amazon Inspector

    Explanation

    Amazon Macie uses managed and custom data identifiers to sample and classify objects in S3, reporting buckets that contain sensitive data such as personal information. It also evaluates bucket-level security and access controls.

  33. Question 33Design Secure Architectures

    A compliance team must prove that CloudTrail log files have not been altered after delivery. Which CloudTrail feature should be enabled?

    1. Log file integrity validation Correct answer.
    2. CloudTrail Insights
    3. An organization trail
    4. CloudTrail Lake event data stores

    Explanation

    Log file integrity validation produces digest files containing hash values for delivered log files, letting you verify that logs were not modified or deleted after delivery. The digests can be validated with the AWS CLI or programmatically.

  34. Question 34Design Secure Architectures

    A security policy requires that every request to an S3 bucket use HTTPS. Which bucket policy element enforces this?

    1. An Allow statement scoped to IpAddress aws:SourceIp 0.0.0.0/0.
    2. A Deny statement with the condition StringEquals s3:x-amz-acl private.
    3. An Allow statement for s3:* with the condition Bool aws:SecureTransport true.
    4. A Deny statement for s3:* with the condition Bool aws:SecureTransport false. Correct answer.

    Explanation

    Because allows are additive and cannot restrict existing permissions, encryption in transit must be enforced with a Deny that triggers when aws:SecureTransport is false. That explicit deny rejects any plain HTTP request, including requests otherwise allowed by other policies.

  35. Question 35Design Secure Architectures

    A regulated workload requires that an object version cannot be deleted or overwritten by any user, including the account root user, for a fixed retention period. Which configuration meets this requirement?

    1. S3 Object Lock in compliance mode with a retention period. Correct answer.
    2. A bucket policy that denies s3:DeleteObjectVersion to all principals.
    3. S3 Object Lock in governance mode with a legal hold applied.
    4. S3 versioning with MFA delete enabled on the bucket.

    Explanation

    Compliance mode retention cannot be shortened or removed by any principal, including the root user, until the retention period expires. Governance mode, by contrast, can be bypassed by users holding s3:BypassGovernanceRetention.

  36. Question 36Design Secure Architectures

    An existing unencrypted Amazon RDS database must be encrypted at rest with minimal downtime. What is the valid approach?

    1. Enable encryption in place by modifying the DB instance's storage settings.
    2. Enable encryption on the DB instance's option group and reboot the instance.
    3. Copy the snapshot with encryption and restore from the copy. Correct answer.
    4. Attach an encrypted EBS volume to the DB instance and move the data files onto it.

    Explanation

    Encryption cannot be turned on for an existing unencrypted RDS instance in place. The supported path is to take a snapshot, copy the snapshot with encryption enabled, and restore a new encrypted instance from that copy.

  37. Question 37Design Secure Architectures

    A security team wants to require IMDSv2 (session-token-based) access to instance metadata on existing EC2 instances. Which action achieves this?

    1. Configure the instance metadata options to require HTTP tokens. Correct answer.
    2. Add an inbound rule on port 80 to the instance's security group.
    3. Enable termination protection on the instances.
    4. Disable the instance profile attached to the instance.

    Explanation

    Setting the instance metadata option HttpTokens to required forces callers to obtain a session token with a PUT request before reading metadata, which mitigates SSRF-based credential theft. The setting can be applied to existing instances without a restart of the workload.

  38. Question 38Design Secure Architectures

    A trust policy must ensure that a role can be assumed only when the caller has authenticated with MFA. Which condition key is used?

    1. aws:MultiFactorAuthPresent Correct answer.
    2. aws:PrincipalOrgID
    3. sts:ExternalId
    4. aws:SourceVpce

    Explanation

    The aws:MultiFactorAuthPresent condition key is true when the calling principal authenticated with an MFA device, so a Bool condition on it in the role trust policy blocks non-MFA callers. The role itself should also require MFA via a policy that denies actions when the key is false.

  39. Question 39Design Secure Architectures

    A company must reach a partner's SaaS API over private IP addresses from its VPC without traversing the internet. Which construct should be used?

    1. A virtual private gateway with a static route to the partner's API.
    2. An interface VPC endpoint for the partner service via AWS PrivateLink. Correct answer.
    3. An internet-facing Network Load Balancer in front of the partner endpoint.
    4. A NAT gateway in a public subnet with a route to the partner's CIDR.

    Explanation

    AWS PrivateLink exposes a service as an elastic network interface endpoint inside your VPC, so traffic to the partner SaaS API stays on the AWS network and never uses the internet. The partner publishes the service and grants your account access.

  40. Question 40Design Secure Architectures

    Which information can be obtained from VPC Flow Logs?

    1. Source and destination IP, ports, and whether traffic was accepted or rejected. Correct answer.
    2. The complete application payload of every packet that traverses the elastic network interface.
    3. The operating system process that opened each connection.
    4. The contents of TLS sessions between clients and servers.

    Explanation

    VPC Flow Logs capture metadata about IP traffic, such as source and destination addresses, ports, protocol, packet and byte counts, and the accept or reject decision. They do not capture payloads or decrypted session contents.

  41. Question 41Design Secure Architectures

    A Lambda function in account A must be invoked by an IAM role in account B. Which change makes this possible?

    1. Attach the AWSLambdaFullAccess managed policy to the caller's role.
    2. Add a trust relationship for account B to the function's execution role.
    3. Add a resource-based policy on the function granting invoke to account B. Correct answer.
    4. Enable the function's reserved concurrency for the calling account.

    Explanation

    Cross-account invocation of a Lambda function is authorized by a resource-based (function) policy that names the trusted account or principal and allows lambda:InvokeFunction. The caller's identity policy alone is insufficient because the resource side must also permit the action.

  42. Question 42Design Secure Architectures

    An Application Load Balancer serves an HTTPS listener. Auditors require that clients negotiate at least TLS 1.2. Which setting enforces this?

    1. A network ACL that blocks TCP port 443 for older clients.
    2. An AWS WAF rule group attached to the load balancer.
    3. The listener security policy set to a TLS 1.2 minimum. Correct answer.
    4. The target group's stickiness attributes.

    Explanation

    The HTTPS listener's security policy defines the minimum protocol version and cipher suites negotiated during the TLS handshake. Selecting a policy whose minimum is TLS 1.2 causes the load balancer to reject older protocol versions.

  43. Question 43Design Secure Architectures

    A CloudFront distribution serves private content from an S3 bucket that must not allow public access. Which mechanism restricts the origin to CloudFront?

    1. A bucket website endpoint used as the distribution origin.
    2. Signed cookies issued for each viewer of the distribution.
    3. An S3 bucket ACL that grants READ permission to the AllUsers group.
    4. An origin access control signing requests from CloudFront. Correct answer.

    Explanation

    Origin access control makes CloudFront sign its requests to the S3 origin, and the bucket policy accepts only that signed distribution, so the bucket can remain fully private. This replaces the older origin access identity mechanism with stronger SigV4 signing.

  44. Question 44Design Secure Architectures

    What does enabling EBS encryption by default for a Region accomplish?

    1. It forces all snapshots shared from other accounts to be re-encrypted with the default key.
    2. New volumes and snapshots are encrypted by default; existing volumes are unchanged. Correct answer.
    3. It immediately encrypts every existing unencrypted EBS volume and snapshot in the Region.
    4. It encrypts volumes in every Region for the account from a single setting.

    Explanation

    The account-level encryption-by-default setting is regional and applies to volumes and snapshots created after it is enabled, using the specified default KMS key. Existing unencrypted volumes are not retroactively encrypted.

  45. Question 45Design Secure Architectures

    A bucket policy should allow access only when requests arrive through a specific S3 gateway VPC endpoint. Which condition key accomplishes this?

    1. aws:SourceIp
    2. aws:PrincipalOrgID
    3. aws:SourceVpce Correct answer.
    4. aws:SecureTransport

    Explanation

    The aws:SourceVpce condition key matches the ID of the VPC endpoint through which the request was made, so a bucket policy can restrict access to that endpoint. Gateway endpoint requests do not originate from a NAT or instance public address, which is why SourceIp is unreliable for this purpose.

  46. Question 46Design Secure Architectures

    A bucket policy must allow read access only to principals in accounts that belong to the company's AWS Organization, including accounts added later. Which condition is most appropriate?

    1. StringEquals on aws:PrincipalOrgID with the organization ID. Correct answer.
    2. StringEquals on aws:PrincipalAccount with a list of the current account IDs.
    3. StringEquals on aws:userid with the organization's root user ID.
    4. ArnLike on aws:PrincipalArn with the organization's management account ARN.

    Explanation

    The aws:PrincipalOrgID global condition key matches the organization of the requesting principal, so the policy automatically covers accounts that join the organization later. An explicit list of account IDs would need to be maintained manually.

  47. Question 47Design Secure Architectures

    An application assumes role A and then uses role A's credentials to assume role B. What is the maximum session duration of the credentials for role B?

    1. The duration configured on role A's trust policy.
    2. 36 hours when the caller originates from an IAM user with long-term credentials.
    3. 12 hours, the same as a directly assumed role.
    4. 1 hour, because role chaining limits the session duration. Correct answer.

    Explanation

    When a role session is used to assume another role, the resulting session is limited to a maximum of one hour regardless of the duration configured on the target role. Directly assumed role sessions can be configured up to twelve hours, but chained sessions cannot.

  48. Question 48Design Secure Architectures

    A company must block outbound DNS lookups for known malicious domains from workloads in a VPC while allowing all other queries. Which solution meets this requirement?

    1. A network ACL that denies UDP port 53 to the internet.
    2. Security groups that allow only approved destinations on port 443.
    3. Route 53 Resolver DNS Firewall with a blocking rule group. Correct answer.
    4. An S3 gateway endpoint policy that denies traffic to the listed domains.

    Explanation

    Route 53 Resolver DNS Firewall inspects queries made through the VPC resolver and can block specific domains or categories while permitting everything else. Domain-level filtering requires a DNS-aware control, which neither security groups nor network ACLs provide.

  49. Question 49Design Secure Architectures

    A team needs an audit record of every object-level read in a specific S3 bucket, including reads performed by principals in other accounts. What must be configured?

    1. S3 data events enabled on the trail for the bucket's ARN prefix. Correct answer.
    2. Management events enabled on the CloudTrail trail for that Region.
    3. A VPC Flow Log on the interface used to reach the bucket.
    4. CloudTrail Insights events enabled on the trail.

    Explanation

    Object-level API activity such as GetObject is recorded only when S3 data events are enabled for the bucket, and the record includes the requesting principal regardless of account. Management events cover control-plane operations like CreateBucket rather than object reads.

  50. Question 50Design Secure Architectures

    In the same account, an IAM role has an identity policy allowing kms:Decrypt on a customer managed key. The key policy contains no statement naming that role. What is the outcome?

    1. The call succeeds because an identity policy is sufficient for any principal in the same account.
    2. The call succeeds only if the role is also listed in a grant created by the key administrator.
    3. The call fails because key policies are never combined with identity policies.
    4. It succeeds only if the key policy delegates to the account, enabling IAM policies. Correct answer.

    Explanation

    Access to a KMS key is always controlled first by the key policy; IAM policies can only contribute when the key policy delegates authority to the account principal, which is present by default. If the key policy has been edited to remove that delegation and does not name the role, the identity policy alone grants nothing.

  51. Question 51Design Resilient Architectures

    Which capability does configuring an Amazon RDS DB instance as a Multi-AZ deployment provide?

    1. Read scaling by serving queries from a second node
    2. Encryption of data at rest using a customer managed key
    3. Automatic failover to a standby in another Availability Zone Correct answer.
    4. Cross-Region disaster recovery with continuous asynchronous replication

    Explanation

    Multi-AZ maintains a synchronous standby copy in a different Availability Zone and promotes it automatically when the primary or its AZ fails. The standby does not serve read traffic, and replication between Regions is handled by cross-Region read replicas. Encryption at rest is configured separately through storage encryption settings.

  52. Question 52Design Resilient Architectures

    A team needs to recover from accidental object overwrites or deletions in an Amazon S3 bucket with the least operational effort. Which feature should they enable?

    1. Enable S3 Transfer Acceleration on the bucket
    2. Add a lifecycle rule that transitions objects to Glacier
    3. Enable default encryption with SSE-KMS on the bucket
    4. Enable S3 Versioning on the bucket Correct answer.

    Explanation

    S3 Versioning retains every prior version of an object, so an overwrite or delete creates a new version or delete marker rather than destroying data. Older versions can be restored by removing the delete marker or copying a prior version forward. Acceleration, lifecycle transitions, and encryption do not preserve prior object states.

  53. Question 53Design Resilient Architectures

    An application runs on EC2 instances behind an Application Load Balancer. Which configuration makes the architecture resilient to the loss of a single Availability Zone?

    1. Enable sticky sessions on the load balancer target group
    2. Replace the Application Load Balancer with a Network Load Balancer
    3. Span the Auto Scaling group across both Availability Zones Correct answer.
    4. Place all instances in one Availability Zone and rely on the load balancer to replicate them

    Explanation

    Resilience to an AZ failure requires running capacity in more than one AZ, which an Auto Scaling group spanning multiple subnets in different AZs provides. The load balancer then routes only to healthy targets in the surviving AZ. Load balancer type and session stickiness affect traffic distribution, not the placement of compute capacity.

  54. Question 54Design Resilient Architectures

    A web tier writes requests to a backend service that occasionally slows down, causing request failures. Which change decouples the tiers most directly?

    1. Enable cross-zone load balancing on the load balancer
    2. Increase the instance size of the backend tier to absorb the load
    3. Add an ElastiCache cluster in front of the database
    4. Insert an Amazon SQS queue between the web tier and the backend Correct answer.

    Explanation

    A queue buffers requests so the web tier can continue accepting traffic while the backend processes at its own pace, absorbing transient slowdowns and spikes. This asynchronous pattern removes the tight runtime coupling between the tiers. Resizing or caching changes capacity and latency but keeps the synchronous dependency.

  55. Question 55Design Resilient Architectures

    Which Amazon Route 53 configuration sends traffic to a secondary endpoint only when the primary endpoint becomes unhealthy?

    1. A geolocation routing policy with a default record for both endpoints
    2. A failover routing policy with a health check on the primary record Correct answer.
    3. A latency routing policy with both records in the same Region
    4. A weighted routing policy with equal weights on both records

    Explanation

    Failover routing designates a primary and a secondary record, and Route 53 serves the secondary only when the health check on the primary reports unhealthy. Weighted routing splits traffic by proportion regardless of health, and latency and geolocation policies choose based on network or user location rather than availability.

  56. Question 56Design Resilient Architectures

    Which approach produces point-in-time backups of an EBS volume that can be restored in a different Availability Zone or Region?

    1. Enable EBS encryption on the volume
    2. Enable detailed monitoring for the volume and create CloudWatch alarms
    3. Create EBS snapshots and restore them in another Availability Zone Correct answer.
    4. Build a RAID 0 array across two volumes in the same Availability Zone

    Explanation

    EBS snapshots are incremental copies stored durably in Amazon S3 and can be used to create volumes in any Availability Zone or Region. Snapshots are independent of the source volume and its AZ, which makes them the standard backup mechanism. Encryption, RAID, and monitoring do not create restorable copies.

  57. Question 57Design Resilient Architectures

    Which Amazon DynamoDB feature allows restoring a table to any point within the previous 35 days?

    1. On-demand backup and restore
    2. Point-in-time recovery Correct answer.
    3. DynamoDB Streams
    4. Global tables

    Explanation

    Point-in-time recovery maintains a continuous backup of table data for 35 days and restores to any second within that window. On-demand backups are user-initiated snapshots that restore only to the moment the backup was taken. Streams capture item-level changes for processing, and global tables replicate data across Regions.

  58. Question 58Design Resilient Architectures

    What does an Amazon RDS read replica provide?

    1. Synchronous replication across multiple Availability Zones
    2. Automatic failover of the primary instance
    3. A standby copy that cannot serve read traffic
    4. Asynchronous replication used to offload read traffic Correct answer.

    Explanation

    A read replica receives asynchronous changes from its source instance and can serve read-only queries, optionally in a different Region. It is not the Multi-AZ standby, does not use synchronous replication, and its promotion to primary is a manual action.

  59. Question 59Design Resilient Architectures

    After an Amazon RDS Multi-AZ failover completes, what must client applications do to reach the database again?

    1. Applications must reconnect through the read replica endpoint
    2. Applications keep using the same DB instance endpoint Correct answer.
    3. Applications must wait for the standby to be demoted to a read replica
    4. Applications must be reconfigured with the new primary IP address

    Explanation

    The DB instance endpoint is a DNS name whose record is repointed to the promoted standby during failover. Applications that reconnect with retry logic to the same endpoint recover without configuration changes, though existing connections are dropped.

  60. Question 60Design Resilient Architectures

    Which prerequisite must be satisfied before enabling S3 Cross-Region Replication from a source bucket to a destination bucket?

    1. Both buckets must use the same AWS KMS key
    2. The destination bucket must be in the same Region as the source bucket
    3. Versioning must be enabled on both the source and destination buckets Correct answer.
    4. The source bucket must have Object Lock enabled

    Explanation

    Cross-Region Replication requires versioning on the source and destination buckets, because replication tracks and copies individual object versions. The destination must be in a different Region by definition, and KMS keys are Region-scoped, so each bucket uses a key in its own Region.

  61. Question 61Design Resilient Architectures

    A company needs a relational database with a cross-Region recovery point objective of about one second and fast promotion of a secondary Region. Which design should be used?

    1. Aurora Global Database with a secondary cluster in another Region Correct answer.
    2. RDS Multi-AZ with automated backups copied to another Region
    3. A MySQL server on EC2 with nightly dumps written to S3 in another Region
    4. A single Aurora cluster with three readers in one Region

    Explanation

    Aurora Global Database replicates storage at the infrastructure level with typical lag under one second and lets a secondary cluster be promoted quickly during a regional event. Multi-AZ plus copied backups and nightly dumps produce recovery points measured in hours, not seconds. Multiple readers in one Region do not protect against a regional failure.

  62. Question 62Design Resilient Architectures

    An application backed by Amazon DynamoDB must accept writes in two Regions simultaneously and keep both copies current. Which design satisfies this?

    1. Create a read replica of the table in the second Region
    2. Use a DynamoDB global table spanning both Regions Correct answer.
    3. Copy nightly backups to the second Region and restore on demand
    4. Enable DynamoDB Accelerator in both Regions

    Explanation

    DynamoDB global tables provide multi-active replication, so each Region can accept reads and writes and changes propagate to the other replicas. Backup copies and caches provide neither write acceptance nor automatic propagation in the second Region.

  63. Question 63Design Resilient Architectures

    Messages in a standard Amazon SQS queue fail processing repeatedly and prevent later work from progressing. Which configuration addresses this?

    1. A dead-letter queue with a redrive policy based on receive count Correct answer.
    2. An increased visibility timeout near the maximum value
    3. A FIFO queue with content-based deduplication enabled for the producers
    4. Long polling with a 20-second wait time

    Explanation

    A redrive policy moves a message to a dead-letter queue once it has been received more than a configured number of times, isolating poison messages for later analysis. This lets the main queue continue processing without manual intervention. Visibility timeout and polling settings change delivery behavior but do not quarantine failing messages.

  64. Question 64Design Resilient Architectures

    An order service must notify an email system, a fraud checker, and an analytics pipeline independently, and each consumer may be unavailable at times. Which design should be used?

    1. A single SQS queue consumed by all three services
    2. Synchronous HTTPS calls from the order service to each consumer
    3. An SNS topic with one SQS queue subscribed per consumer Correct answer.
    4. Each consumer polls the orders table on a scheduled Lambda function

    Explanation

    Publishing to an SNS topic pushes a copy of each message to every subscribed queue, so each consumer processes independently and buffers work while it is down. Sharing one queue makes consumers compete for messages rather than each receiving a copy. Synchronous calls reintroduce tight coupling.

  65. Question 65Design Resilient Architectures

    An Auto Scaling group has instances spread across three Availability Zones. After one zone recovers from an outage, the distribution is uneven. Which action restores balance?

    1. Enable termination protection on the Auto Scaling group
    2. Raise the maximum size to three times the desired capacity
    3. Manually terminate instances in the healthy Availability Zones
    4. Enable the Availability Zone rebalance option on the group Correct answer.

    Explanation

    Availability Zone rebalancing lets the group redistribute instances across its enabled zones when capacity becomes uneven, for example after a zone recovers. Termination protection and a larger maximum size change scaling constraints but not placement.

  66. Question 66Design Resilient Architectures

    Multiple EC2 instances in different Availability Zones need concurrent read-write access to the same POSIX-compliant file system. Which storage service should be used?

    1. An Amazon S3 bucket mounted as a file system
    2. Amazon EFS Correct answer.
    3. An EBS io2 volume with Multi-Attach enabled
    4. Instance store volumes on each instance

    Explanation

    Amazon EFS is a regional, elastic NFS file system that many instances in different Availability Zones can mount and write to concurrently. EBS Multi-Attach is limited to instances in a single Availability Zone, and instance store is ephemeral and local to one host.

  67. Question 67Design Resilient Architectures

    A company must retain EBS, RDS, and DynamoDB backups in a second Region for seven years under one centrally managed policy. Which service fits best?

    1. AWS DataSync tasks scheduled between the two Regions
    2. Native per-service snapshot schedules configured in each Region
    3. AWS Backup with a cross-Region copy and a vault lock retention policy Correct answer.
    4. Amazon S3 Lifecycle rules on a bucket in the second Region

    Explanation

    AWS Backup provides a single policy-driven framework across supported services, including cross-Region copy and vault lock for immutable retention. Managing each service's native snapshots separately multiplies operational overhead and lacks a unified retention control.

  68. Question 68Design Resilient Architectures

    Instances in private subnets across three Availability Zones need outbound internet access that survives the loss of one Availability Zone. Which design meets this requirement?

    1. A single NAT gateway in one Availability Zone shared by all of the route tables
    2. One NAT gateway per Availability Zone, referenced by that zone's route table Correct answer.
    3. An internet gateway attached to each private subnet
    4. A VPC peering connection to a public subnet in another VPC

    Explanation

    A NAT gateway is confined to one Availability Zone, so deploying one per zone and routing each private subnet to its local gateway prevents a single zone failure from breaking egress. A shared gateway becomes a single point of failure. Internet gateways attach at the VPC level and do not by themselves provide private-subnet egress.

  69. Question 69Design Resilient Architectures

    A disaster recovery plan specifies an RPO of 15 minutes and an RTO of 1 hour. What do these values mean?

    1. Failover must be tested every 15 minutes and completed within 1 hour
    2. At most 15 minutes of downtime and up to 1 hour of data loss
    3. Backups run every hour and are retained for 15 minutes
    4. Up to 15 minutes of data loss and service restored within 1 hour Correct answer.

    Explanation

    Recovery point objective defines the maximum acceptable amount of data loss measured in time, while recovery time objective defines the maximum acceptable time to restore service. An RPO of 15 minutes means recovery points must be no more than 15 minutes old, and an RTO of 1 hour means service must resume within an hour.

  70. Question 70Design Resilient Architectures

    A Redis cache must continue serving after a node failure with no manual intervention. Which configuration should be used?

    1. A single Redis node with daily snapshots
    2. Cluster mode disabled with no replicas
    3. Multi-AZ with automatic failover enabled Correct answer.
    4. Memcached with Auto Discovery enabled on the client

    Explanation

    Multi-AZ with automatic failover keeps a replica in another Availability Zone and promotes it when the primary fails, restoring the same endpoint without operator action. A single node or a replication group without failover requires manual recovery, and Memcached does not support replication at all.

  71. Question 71Design Resilient Architectures

    A disaster recovery plan keeps a scaled-down but fully functional copy of production running in a second Region that can be scaled to full capacity quickly. Which strategy does this describe?

    1. Backup and restore
    2. Pilot light
    3. Multi-site active-active
    4. Warm standby Correct answer.

    Explanation

    Warm standby runs a complete but reduced-capacity version of the workload in the recovery Region, so failover is a scaling step rather than a build step. Pilot light runs only core components such as replicated data, with application servers created during recovery. Backup and restore builds everything after the event, and multi-site serves traffic from both Regions at full capacity.

  72. Question 72Design Resilient Architectures

    Which Amazon S3 capability provides an SLA-backed commitment that most objects replicate to another Region within 15 minutes?

    1. S3 Transfer Acceleration on both buckets
    2. S3 Replication Time Control Correct answer.
    3. S3 Cross-Region Replication with priority replication enabled
    4. S3 Multi-Region Access Points

    Explanation

    Replication Time Control adds a 15-minute replication time objective backed by an SLA and publishes replication metrics and notifications. Priority replication changes the order in which objects are replicated but carries no time commitment, and acceleration and Multi-Region Access Points address upload speed and routing rather than replication timing.

  73. Question 73Design Resilient Architectures

    A workload running on a standard RDS Multi-AZ instance deployment needs faster failover and wants standby capacity to serve read traffic. Which option should be chosen?

    1. Use an RDS Multi-AZ DB cluster with two readable standbys Correct answer.
    2. Convert the instance to Single-AZ and add five read replicas
    3. Enable Performance Insights on the existing instance
    4. Increase the instance class of the primary DB instance

    Explanation

    An RDS Multi-AZ DB cluster runs three instances across three Availability Zones, exposing two readable standbys and typically failing over much faster than a Multi-AZ instance deployment. Standbys in the instance deployment are not readable and are not exposed as endpoints.

  74. Question 74Design Resilient Architectures

    A company needs to shift traffic between two Regions by using an operator-controlled switch that remains available during a regional impairment. Which solution fits?

    1. Route 53 latency records with a long TTL
    2. An Application Load Balancer with cross-zone load balancing
    3. Route 53 Application Recovery Controller routing controls Correct answer.
    4. Route 53 weighted records updated by a Lambda function

    Explanation

    Route 53 Application Recovery Controller provides routing controls built on a highly available data plane that spans multiple Regions, so the switch can be flipped even when one Region's control plane is impaired. Updating DNS records with a Lambda function depends on the impaired Region and on DNS propagation delays.

  75. Question 75Design Resilient Architectures

    A team wants to confirm that multi-AZ failover behaves as designed before relying on it in production. Which approach is appropriate?

    1. Run an AWS Fault Injection Service experiment in a test environment Correct answer.
    2. Lower the Route 53 record TTL to 0 seconds to speed failover
    3. Disable load balancer health checks to force a failover
    4. Wait for a real Availability Zone failure and observe the results

    Explanation

    AWS Fault Injection Service runs controlled experiments such as disrupting a single Availability Zone so the team can observe failover behavior under realistic conditions without waiting for an actual outage. Disabling health checks alters the system under test and does not validate the failover design.

  76. Question 76Design Resilient Architectures

    A team runs a small number of critical EC2 instances that must not share the same underlying hardware or rack with one another. Which placement group strategy should they use?

    1. Dedicated Host
    2. Cluster placement group
    3. Spread placement group Correct answer.
    4. Partition placement group

    Explanation

    A spread placement group places instances on distinct underlying hardware to reduce the risk of correlated failures. Cluster placement groups pack instances closely for low latency, while partition groups are intended for large distributed workloads.

  77. Question 77Design Resilient Architectures

    An application writes session data to the instance store of an EC2 instance. What happens to that data if the instance is stopped and started?

    1. It is lost because instance store is ephemeral. Correct answer.
    2. It is retained for seven days before being deleted.
    3. It is retained because instance store is backed by Amazon EBS.
    4. It is automatically replicated to another Availability Zone.

    Explanation

    Instance store volumes are physically attached to the host and provide temporary block storage. Stopping and starting an instance normally moves it to different hardware, so the data on instance store is lost.

  78. Question 78Design Resilient Architectures

    A primary and secondary web endpoint run in different Regions. The company wants Route 53 to send traffic to the secondary only when the primary fails its health check. Which routing policy meets this requirement?

    1. Failover routing Correct answer.
    2. Geolocation routing
    3. Weighted routing
    4. Latency-based routing

    Explanation

    Failover routing uses a health check on the primary record and automatically answers with the secondary record when the primary is unhealthy. It is designed specifically for active-passive failover between endpoints.

  79. Question 79Design Resilient Architectures

    An application overwrites objects in an Amazon S3 bucket by mistake. The team wants to be able to recover the previous version of each object after an overwrite. What should they enable?

    1. S3 Versioning Correct answer.
    2. S3 Object Lock in governance mode
    3. S3 Transfer Acceleration
    4. S3 Cross-Region Replication

    Explanation

    S3 Versioning keeps multiple variants of an object in the same bucket, so an overwrite creates a new version while the prior version remains recoverable. Object Lock and replication serve different purposes and do not provide version recovery by themselves.

  80. Question 80Design Resilient Architectures

    What is the maximum retention period for automated backups of an Amazon RDS database instance?

    1. 365 days
    2. 35 days Correct answer.
    3. 90 days
    4. 7 days

    Explanation

    Amazon RDS automated backups can be retained for up to 35 days. Manual snapshots are separate and can be kept for as long as needed.

  81. Question 81Design Resilient Architectures

    An EC2 instance runs a stateful application and must be recovered automatically when AWS detects a system status check failure. Which feature should be configured?

    1. EC2 auto recovery with a CloudWatch alarm on the StatusCheckFailed_System metric Correct answer.
    2. A Spot Fleet request with capacity-optimized allocation
    3. An Elastic Load Balancer health check on the instance
    4. An Auto Scaling group with a minimum size of zero

    Explanation

    EC2 auto recovery uses a CloudWatch alarm on the system status check to recover the instance on new hardware while preserving its instance ID and attached EBS volumes. An Auto Scaling group replaces instances but is a different recovery model.

  82. Question 82Design Resilient Architectures

    Which disaster recovery strategy has the lowest cost but the longest recovery time?

    1. Pilot light
    2. Backup and restore Correct answer.
    3. Multi-site active-active
    4. Warm standby

    Explanation

    Backup and restore keeps only backups and rebuilds infrastructure during recovery, which is inexpensive but slow. The other strategies maintain more running capacity and therefore cost more while recovering faster.

  83. Question 83Design Resilient Architectures

    Which AWS service provides managed DDoS protection with access to the AWS DDoS Response Team and cost protection for scaling charges during an attack?

    1. AWS WAF
    2. AWS Shield Advanced Correct answer.
    3. Amazon Inspector
    4. Amazon GuardDuty

    Explanation

    AWS Shield Advanced provides enhanced DDoS protection, access to the AWS DDoS Response Team, and cost protection for eligible scaling charges during an attack. AWS WAF provides web request filtering but not the same DDoS response and cost-protection features.

  84. Question 84Design Resilient Architectures

    CloudFront serves content from a primary S3 bucket. The team wants CloudFront to automatically serve from a secondary bucket when the primary returns 5xx errors. What should they configure?

    1. A Route 53 failover record that points to the secondary bucket
    2. An origin group with the primary as the primary origin and the secondary as the failover origin Correct answer.
    3. A Lambda@Edge function that rewrites all requests to the secondary bucket
    4. A cache policy with a longer minimum TTL

    Explanation

    CloudFront origin groups define a primary and a secondary origin, and CloudFront fails over based on configurable error responses. This keeps failover within the distribution without requiring DNS changes.

  85. Question 85Design Resilient Architectures

    An Application Load Balancer has targets in two Availability Zones. One zone has more healthy targets than the other, and the team wants requests distributed evenly across all healthy targets regardless of zone. What should they enable?

    1. Slow start
    2. Cross-zone load balancing Correct answer.
    3. Connection draining
    4. Sticky sessions

    Explanation

    Cross-zone load balancing lets each load balancer node distribute traffic evenly across healthy targets in all enabled Availability Zones. Without it, each zone's load balancer nodes primarily route to targets in their own zone.

  86. Question 86Design Resilient Architectures

    An Auto Scaling group replaces instances only when EC2 status checks fail. The application behind an ALB becomes unhealthy while the instance still passes EC2 status checks. What should the team change?

    1. Disable instance scale-in protection
    2. Increase the group's maximum size
    3. Change the termination policy to OldestInstance
    4. Set the health check type to ELB Correct answer.

    Explanation

    When the health check type is ELB, the Auto Scaling group also considers load balancer health checks and can replace instances that fail application-level checks. EC2 health checks alone detect only instance-level failures.

  87. Question 87Design Resilient Architectures

    Regulated data in Amazon S3 must be immutable for a fixed period, and no user, including the root user, should be able to delete or overwrite it during that period. What should be used?

    1. S3 Object Lock in compliance mode Correct answer.
    2. S3 Cross-Region Replication
    3. A bucket policy that denies s3:DeleteObject
    4. S3 Versioning with MFA delete

    Explanation

    S3 Object Lock in compliance mode enforces a retention period that even the root user cannot shorten or bypass. Governance mode and ordinary bucket policies can be changed by sufficiently privileged users.

  88. Question 88Design Resilient Architectures

    An application opens many short-lived connections to an RDS Multi-AZ database. During a failover, clients wait for connections to time out before reconnecting. Which service can reduce this interruption by pooling and holding connections?

    1. Amazon Route 53
    2. Amazon RDS Proxy Correct answer.
    3. Amazon ElastiCache
    4. AWS Global Accelerator

    Explanation

    Amazon RDS Proxy maintains a connection pool and can route connections to the new primary after a failover, reducing client connection churn and timeouts. It is designed for database connection management rather than caching or network routing.

  89. Question 89Design Resilient Architectures

    An application is deployed in two Regions behind Application Load Balancers. Users must be routed to the nearest healthy Regional endpoint with fast failover, using static anycast IP addresses. Which service should be used?

    1. Network Load Balancer with sticky sessions
    2. Amazon CloudFront with a cache policy
    3. Amazon Route 53 latency records
    4. AWS Global Accelerator Correct answer.

    Explanation

    AWS Global Accelerator provides static anycast IP addresses and routes users to the nearest healthy Regional endpoint with fast failover. Route 53 latency records use DNS and do not provide static anycast IP addresses.

  90. Question 90Design Resilient Architectures

    A company wants continuous block-level replication of on-premises servers to AWS so that it can launch recovery instances within minutes after a disaster, without maintaining a warm standby fleet. Which service is designed for this?

    1. AWS Elastic Disaster Recovery Correct answer.
    2. AWS Storage Gateway
    3. AWS Backup
    4. AWS DataSync

    Explanation

    AWS Elastic Disaster Recovery continuously replicates source servers into a staging area in AWS and can launch recovery instances on demand. It is intended for low-cost disaster recovery without a continuously running standby fleet.

  91. Question 91Design Resilient Architectures

    A disaster recovery plan keeps a minimal core environment running in a second Region, with data replicated and servers stopped. During a disaster, the team scales up the environment. Which strategy is this?

    1. Backup and restore
    2. Pilot light Correct answer.
    3. Warm standby
    4. Multi-site active-active

    Explanation

    A pilot light strategy keeps a minimal core environment and replicated data running, then scales up the remaining resources during recovery. Warm standby maintains a larger, fully functional scaled-down copy.

  92. Question 92Design Resilient Architectures

    A security team must prevent deletion of recovery points in an AWS Backup vault, even by administrators, for a fixed retention period. What should they configure?

    1. AWS Backup Vault Lock in compliance mode Correct answer.
    2. AWS Backup cross-Region copy
    3. An IAM policy that denies backup:DeleteRecoveryPoint
    4. S3 Object Lock on the backup bucket

    Explanation

    AWS Backup Vault Lock in compliance mode enforces immutability for the configured retention period and cannot be removed by administrators during that period. An IAM policy alone can be changed by a sufficiently privileged administrator.

  93. Question 93Design Resilient Architectures

    A company needs a highly available private connection to AWS from a single data center. Which design provides redundancy if one Direct Connect connection or device fails?

    1. A single Direct Connect connection with a larger port size
    2. A single Direct Connect connection with a VPN backup over the internet
    3. One Direct Connect connection and a Route 53 health check
    4. Two Direct Connect connections terminating at two different Direct Connect locations Correct answer.

    Explanation

    Two Direct Connect connections terminating at different Direct Connect locations provide redundancy at the connection and location level. A single connection, even with a larger port, remains a single point of failure.

  94. Question 94Design Resilient Architectures

    An application must be able to replay past events to a new consumer after a deployment error. Which EventBridge feature supports storing events and replaying them later?

    1. EventBridge API destinations
    2. EventBridge archive and replay Correct answer.
    3. EventBridge schema registry
    4. EventBridge global endpoints

    Explanation

    EventBridge archive and replay stores events for a retention period and lets you replay them to an event bus or target later. This supports recovery after a consumer deployment error.

  95. Question 95Design Resilient Architectures

    An application requires a load balancer endpoint with static IP addresses that clients can allowlist and that remains available if one Availability Zone fails. Which load balancer type should be used?

    1. Classic Load Balancer with sticky sessions
    2. Application Load Balancer with cross-zone load balancing
    3. Gateway Load Balancer
    4. Network Load Balancer with one Elastic IP per Availability Zone Correct answer.

    Explanation

    A Network Load Balancer can have one Elastic IP address per Availability Zone, giving clients stable allowlistable addresses. If one zone fails, the load balancer can continue serving through the remaining zones.

  96. Question 96Design Resilient Architectures

    A company needs a single global endpoint for an S3 bucket replicated to multiple Regions, with automatic failover to the nearest healthy copy. Which feature provides this?

    1. Route 53 latency-based records for each bucket
    2. S3 Multi-Region Access Point Correct answer.
    3. CloudFront with an origin group
    4. S3 Cross-Region Replication with a bucket alias

    Explanation

    S3 Multi-Region Access Points provide a single global endpoint that routes requests across replicated buckets and can fail over to a healthy copy. Cross-Region Replication alone replicates data but does not provide the global endpoint.

  97. Question 97Design Resilient Architectures

    A Route 53 health check must report unhealthy only when more than 20 percent of its child health checks are unhealthy. Which feature should be used?

    1. Route 53 failover routing with a primary record
    2. Route 53 private hosted zone
    3. Route 53 calculated health check Correct answer.
    4. Route 53 multivalue answer routing

    Explanation

    A calculated health check combines the status of child health checks and can trigger based on a threshold, such as a percentage of unhealthy children. This avoids treating a single child failure as a full outage.

  98. Question 98Design Resilient Architectures

    An Aurora MySQL cluster needs to be rewound to a point in time within its retention period without restoring a new cluster or changing the endpoint. Which feature supports this?

    1. Aurora Global Database
    2. Aurora automated backups restore
    3. Aurora read replica promotion
    4. Aurora backtrack Correct answer.

    Explanation

    Aurora backtrack rewinds an Aurora MySQL cluster in place to an earlier point in time without creating a new cluster. Restoring from backups creates a new cluster and requires endpoint changes.

  99. Question 99Design Resilient Architectures

    A company wants to assess its workloads against resilience best practices and identify availability weaknesses across accounts. Which service should be used?

    1. AWS Config
    2. AWS Resilience Hub Correct answer.
    3. AWS Well-Architected Tool
    4. AWS Trusted Advisor

    Explanation

    AWS Resilience Hub assesses applications against resilience policies and best practices, helping identify weaknesses relative to recovery objectives. It is specifically focused on application resilience rather than general configuration compliance.

  100. Question 100Design Resilient Architectures

    A team wants to verify that AWS Backup recovery points can actually restore within the required RTO. Which AWS Backup capability should they use?

    1. AWS Backup cross-account copy
    2. AWS Backup restore testing Correct answer.
    3. AWS Backup Audit Manager
    4. AWS Backup Vault Lock

    Explanation

    AWS Backup restore testing performs scheduled restore jobs from recovery points and helps validate that backups meet recovery time objectives. Audit Manager, Vault Lock, and cross-account copy support governance and durability but do not validate restore timing.

  101. Question 101Design High-Performing Architectures

    A team needs an in-memory cache layer that can survive a node failure without losing the cached dataset and that supports persistent backups. Which engine should they choose?

    1. ElastiCache for Redis Correct answer.
    2. DynamoDB Accelerator (DAX)
    3. S3 Intelligent-Tiering
    4. ElastiCache for Memcached

    Explanation

    ElastiCache for Redis supports replication, Multi-AZ with automatic failover, and persistence through backups and append-only files. Memcached is a pure transient cache with no replication or persistence. DAX accelerates DynamoDB only, and S3 tiering is object storage, not a cache.

  102. Question 102Design High-Performing Architectures

    A web application stores static images in an S3 bucket in a single Region and serves users worldwide who report slow image loads. Which service improves load times by caching content at edge locations close to users?

    1. Amazon Route 53 latency routing
    2. AWS Global Accelerator
    3. AWS Direct Connect
    4. Amazon CloudFront Correct answer.

    Explanation

    CloudFront is a content delivery network that caches objects at edge locations, so repeat requests are served from a location near the user instead of the origin Region. Global Accelerator optimizes network paths for dynamic traffic but does not cache content. Direct Connect is a private network link, not a caching layer.

  103. Question 103Design High-Performing Architectures

    Users on other continents upload large files to an S3 bucket in us-east-1 and throughput is poor because of long-distance transfers. Which S3 feature speeds up these uploads by entering the AWS network at an edge location?

    1. S3 Transfer Acceleration Correct answer.
    2. S3 Object Lock
    3. S3 Requester Pays
    4. S3 Cross-Region Replication

    Explanation

    Transfer Acceleration uses CloudFront edge locations to accept uploads close to the user and then moves the data over the optimized AWS backbone to the destination bucket. Replication copies objects after they arrive, Requester Pays shifts billing, and Object Lock is a retention control.

  104. Question 104Design High-Performing Architectures

    A MySQL RDS instance is saturated by a reporting workload that only runs read queries. What is the most direct way to offload those reads?

    1. Enable a Multi-AZ standby instance
    2. Add RDS read replicas for reports Correct answer.
    3. Increase the backup retention period
    4. Turn on Performance Insights for the DB

    Explanation

    Read replicas maintain their own copy of the data and can serve read-only query traffic, removing load from the primary instance. A Multi-AZ standby exists only for failover and cannot be read from. Retention settings and Performance Insights do not add read capacity.

  105. Question 105Design High-Performing Architectures

    An application must deliver each order event to a payment queue, a fraud queue, and an email queue using a single publish call. Which design achieves this?

    1. One SQS queue read by three consumers
    2. Three SQS queues written by the producer
    3. A DynamoDB table scanned by each consumer
    4. SNS topic with three SQS subscriptions Correct answer.

    Explanation

    Publishing once to an SNS topic that has three SQS subscriptions delivers an identical message to each queue in a fan-out pattern. A single queue with multiple consumers divides messages rather than duplicating them. Writing to three queues separately requires three publish calls and couples the producer to each consumer.

  106. Question 106Design High-Performing Architectures

    Which statement correctly describes EC2 instance store volumes?

    1. Data is lost when the instance is stopped Correct answer.
    2. They can be detached and moved to another instance
    3. They are backed up automatically to S3
    4. They are replicated across Availability Zones

    Explanation

    Instance store volumes are physically attached to the host and provide ephemeral block storage; data survives a reboot but is lost when the instance is stopped, terminated, or fails. They are not replicated, cannot be detached and reattached elsewhere, and have no automatic backup.

  107. Question 107Design High-Performing Architectures

    An application uploads 8 GB files to S3 over an unreliable network, and transfers frequently fail near completion. Which change most improves both throughput and resilience?

    1. Enable S3 Object Lock in governance mode
    2. Turn on S3 versioning for the bucket
    3. Upload the files through CloudFront
    4. Use S3 multipart upload in parallel Correct answer.

    Explanation

    Multipart upload splits a large object into parts that are transferred in parallel and can be retried individually, which raises throughput and avoids restarting the whole transfer after a failure. CloudFront, Object Lock, and versioning do not change how the upload itself is performed.

  108. Question 108Design High-Performing Architectures

    An application is deployed in three Regions, and each user must be sent to the Region that gives them the lowest network latency. Which Route 53 routing policy fits?

    1. Weighted routing
    2. Latency-based routing Correct answer.
    3. Multivalue answer routing
    4. Failover routing

    Explanation

    Latency-based routing resolves queries to the Region whose measured latency between the resolver and each record's endpoint is lowest. Weighted routing splits traffic by fixed proportions, failover routing is active-passive, and multivalue answer routing returns multiple healthy records without latency measurement.

  109. Question 109Design High-Performing Architectures

    A workload needs 3,000 IOPS and 500 MB/s of throughput on a 1 TiB EBS volume, and the team wants to provision IOPS and throughput independently of volume size at the lowest cost. Which volume type should they choose?

    1. General Purpose SSD (gp3) Correct answer.
    2. Throughput Optimized HDD (st1)
    3. Provisioned IOPS SSD (io2)
    4. General Purpose SSD (gp2)

    Explanation

    gp3 includes a 3,000 IOPS and 125 MB/s baseline for any volume size and lets you provision additional IOPS and throughput separately. gp2 scales performance with volume size through burst credits, and st1 is a throughput-oriented HDD that cannot deliver these IOPS. io2 can meet the numbers but costs more than needed.

  110. Question 110Design High-Performing Architectures

    Six critical EC2 instances must be placed on distinct underlying hardware to limit correlated failures, and the design may use at most seven running instances per Availability Zone per group. Which placement strategy meets this?

    1. Spread placement group Correct answer.
    2. Cluster placement group
    3. Dedicated Hosts with host affinity
    4. Partition placement group

    Explanation

    A spread placement group places each instance on distinct rack-level hardware and allows up to seven running instances per Availability Zone per group. Cluster groups pack instances together for low latency, partition groups isolate groups of instances into partitions, and Dedicated Hosts address licensing rather than rack separation.

  111. Question 111Design High-Performing Architectures

    A DynamoDB-backed API has read latency spikes during traffic bursts, and the team wants microsecond read latency for repeated item reads with minimal application changes. Which service helps?

    1. DynamoDB Accelerator (DAX) Correct answer.
    2. DynamoDB global tables
    3. DynamoDB on-demand capacity mode
    4. ElastiCache for Memcached

    Explanation

    DAX is a fully managed in-memory cache for DynamoDB that serves eventually consistent reads in microseconds and works through a drop-in DAX client for the existing SDK calls. Global tables replicate data across Regions for locality and availability, on-demand mode changes capacity billing, and Memcached would require custom caching logic.

  112. Question 112Design High-Performing Architectures

    A company needs a relational database with a secondary Region that serves read traffic with typical replication lag under one second and can be promoted quickly after a regional failure. Which design meets this?

    1. Aurora with a Multi-AZ deployment only
    2. RDS Multi-AZ with cross-Region backup copies
    3. Aurora Global Database Correct answer.
    4. RDS for MySQL with a cross-Region read replica

    Explanation

    Aurora Global Database uses a dedicated replication infrastructure that typically propagates changes to secondary Regions in well under one second and supports fast managed promotion. Standard cross-Region read replicas use asynchronous binary log replication with higher lag, and Multi-AZ within one Region does not provide cross-Region reads.

  113. Question 113Design High-Performing Architectures

    A TCP-based gaming service needs a static IP address per Availability Zone and must sustain millions of short-lived connections with very low latency. Which load balancer fits?

    1. Network Load Balancer Correct answer.
    2. Gateway Load Balancer
    3. Classic Load Balancer
    4. Application Load Balancer with sticky sessions

    Explanation

    Network Load Balancer operates at layer 4, handles extreme connection volumes with very low latency, and supports one static IP address per Availability Zone. Application Load Balancer is layer 7 for HTTP and HTTPS, and Gateway Load Balancer is for transparent inspection appliances.

  114. Question 114Design High-Performing Architectures

    An Auto Scaling group must keep average CPU utilization near 50% while traffic varies unpredictably throughout the day. Which scaling configuration is most appropriate?

    1. Simple scaling with a long cooldown
    2. Target tracking scaling policy Correct answer.
    3. Scheduled scaling actions
    4. Manual capacity adjustments

    Explanation

    A target tracking policy adjusts desired capacity automatically to hold a chosen metric at a target value, which suits unpredictable load. Scheduled scaling requires a predictable time pattern, and simple scaling with a long cooldown reacts slowly and in steps rather than continuously.

  115. Question 115Design High-Performing Architectures

    Multiple Linux EC2 instances spread across two Availability Zones must share a POSIX-compliant file system that grows automatically with data volume. Which storage service should be used?

    1. Amazon EBS with Multi-Attach enabled
    2. Amazon Elastic File System Correct answer.
    3. Instance store volumes shared by the group
    4. Amazon S3 mounted with a file gateway

    Explanation

    Amazon EFS is a managed NFS file system that many Linux instances in different Availability Zones can mount concurrently and that scales storage elastically. EBS Multi-Attach is limited to a single Availability Zone and requires a cluster-aware file system, and instance store cannot be shared between instances.

  116. Question 116Design High-Performing Architectures

    A company runs a UDP service in two Regions and wants deterministic, low-latency entry points with static anycast IP addresses. Which service fits?

    1. Route 53 latency-based routing
    2. Amazon CloudFront
    3. AWS Direct Connect gateway
    4. AWS Global Accelerator Correct answer.

    Explanation

    Global Accelerator provides static anycast IP addresses and routes UDP and TCP traffic onto the AWS global network to the optimal Regional endpoint. CloudFront serves HTTP and HTTPS content caching, latency routing only affects DNS resolution and depends on resolver caching, and a Direct Connect gateway links private networks rather than public clients.

  117. Question 117Design High-Performing Architectures

    A latency-sensitive Lambda function shows frequent cold starts after scale-out events on predictable traffic. Which configuration reduces cold-start latency for that traffic?

    1. Increasing function memory to 10 GB
    2. Provisioned concurrency on a published alias Correct answer.
    3. A larger deployment package with bundled libraries
    4. Reserved concurrency at the expected peak

    Explanation

    Provisioned concurrency initializes execution environments in advance and keeps them ready, so invocations on that alias skip initialization. Reserved concurrency caps and guarantees capacity but still allows cold starts, and larger packages and higher memory do not eliminate initialization delay.

  118. Question 118Design High-Performing Architectures

    A REST API on API Gateway repeatedly issues identical GET requests that hit the backend database. Which feature reduces backend load without changing the backend?

    1. API Gateway request validation
    2. API Gateway stage caching Correct answer.
    3. API Gateway usage plans
    4. API Gateway throttling limits

    Explanation

    Stage-level caching stores responses for a configured TTL so repeated identical requests are served from the cache instead of reaching the integration. Usage plans meter API keys, request validation rejects malformed payloads, and throttling limits request rates but still forwards allowed calls to the backend.

  119. Question 119Design High-Performing Architectures

    A Kinesis Data Stream must ingest 2 MB/s of records at 1,000 records per second. What is the minimum number of shards required?

    1. 2 shards Correct answer.
    2. 3 shards
    3. 1 shard
    4. 4 shards

    Explanation

    Each shard supports up to 1 MB/s or 1,000 records/s of writes. The bandwidth requirement of 2 MB/s needs two shards, while the record rate fits within one, so the bandwidth constraint determines the answer.

  120. Question 120Design High-Performing Architectures

    EC2 instances in a private subnet call S3 frequently, and the team wants traffic to stay on the AWS network without NAT gateway charges or additional per-hour endpoint costs. Which should be configured?

    1. An interface VPC endpoint with private DNS
    2. A gateway VPC endpoint for Amazon S3 Correct answer.
    3. AWS PrivateLink behind a Network Load Balancer
    4. A NAT gateway in each Availability Zone

    Explanation

    A gateway VPC endpoint for S3 routes subnet traffic privately to S3 through route table entries at no hourly or data processing charge. Interface endpoints keep traffic private too but bill an hourly and per-GB charge, and NAT gateways incur both hourly and data processing costs.

  121. Question 121Design High-Performing Architectures

    A genomics pipeline on hundreds of EC2 instances needs a scratch file system delivering hundreds of GB/s of throughput that is linked to an S3 bucket for input and output. Which service fits?

    1. Amazon EFS with Max I/O mode
    2. Amazon EBS io2 Block Express
    3. Amazon FSx for Lustre Correct answer.
    4. Amazon FSx for Windows File Server

    Explanation

    FSx for Lustre is a high-performance parallel file system built for compute-heavy workloads and can link a file system to an S3 bucket for transparent input and output. EFS provides shared NFS storage but not the aggregate throughput of a parallel file system, and FSx for Windows File Server targets SMB workloads.

  122. Question 122Design High-Performing Architectures

    An application writes 5,000 new objects per second into one S3 bucket with keys that all begin with the same date prefix. What is the recommended performance practice?

    1. Add a CloudFront distribution in front of S3
    2. Enable S3 Transfer Acceleration on the bucket
    3. No key-prefix changes are needed Correct answer.
    4. Distribute keys across many hash prefixes

    Explanation

    S3 automatically partitions buckets to handle high request rates, so randomizing key prefixes is no longer required for write throughput. Transfer Acceleration and CloudFront address network distance for transfers and reads, not the request-rate capacity of the bucket itself.

  123. Question 123Design High-Performing Architectures

    A DynamoDB table is throttled because its partition key has very few distinct values that are each extremely popular. Which change most directly resolves the uneven distribution of writes across partitions?

    1. Create a global secondary index on the same key
    2. Increase provisioned write capacity units
    3. Add a random suffix to the partition key Correct answer.
    4. Switch the table to on-demand capacity mode

    Explanation

    A low-cardinality key concentrates items into a small number of partitions, so a few partitions absorb most traffic and throttle. Write sharding with a random suffix spreads items, and therefore writes, across many partitions. Raising capacity or changing the capacity mode does not redistribute items, and an index on the same key inherits the same imbalance.

  124. Question 124Design High-Performing Architectures

    A site sells access to thousands of video files. After a user authenticates, that session must be able to fetch any of the files for a limited time, and issuing a separate URL for every file is impractical. Which CloudFront mechanism fits?

    1. CloudFront signed cookies Correct answer.
    2. An origin access identity with a referer check
    3. CloudFront signed URLs with short expirations
    4. Field-level encryption on the distribution

    Explanation

    Signed cookies let CloudFront authorize a viewer for multiple restricted objects without changing each URL, which suits a session that streams many files. Signed URLs authorize one specific resource at a time, and origin access identity protects the S3 origin from direct access rather than authorizing individual viewers.

  125. Question 125Design High-Performing Architectures

    An application must fail over automatically from a primary Region to a standby Region within about a minute without changing the hostname clients use, and the standby must receive traffic only when the primary is unhealthy. Which Route 53 design fits?

    1. Latency-based routing across both Regions
    2. Weighted routing with equal weights
    3. Multivalue answer routing without health checks
    4. Failover routing with a primary health check Correct answer.

    Explanation

    Failover routing uses a health check on the primary record and automatically answers with the secondary record when the primary is unhealthy, keeping the same hostname with a short TTL. Weighted and latency policies send traffic to both Regions simultaneously, and multivalue answers without health checks can return an unhealthy endpoint.

  126. Question 126Design High-Performing Architectures

    An e-commerce site runs on an Application Load Balancer. Requests to /api/* must go to a service target group, while requests to /static/* must go to a different target group. Which ALB feature should be used?

    1. Listener rules that route by host header
    2. Listener rules that route by path pattern Correct answer.
    3. Cross-zone load balancing across Availability Zones
    4. Target group stickiness with cookie duration

    Explanation

    Application Load Balancer listener rules can evaluate the request path and forward matching requests to a specific target group. Host-header rules route by domain name, while stickiness and cross-zone balancing affect session affinity and target distribution rather than path selection.

  127. Question 127Design High-Performing Architectures

    A company wants 10% of production traffic to reach a new application version while 90% continues to the current version, using a single DNS name. Which Route 53 routing policy fits this requirement?

    1. Weighted routing Correct answer.
    2. Geolocation routing
    3. Latency-based routing
    4. Simple routing

    Explanation

    Weighted routing assigns numeric weights to records so that DNS responses are returned in proportion to those weights. This allows a controlled percentage of traffic to reach a new version while most traffic continues to the existing version.

  128. Question 128Design High-Performing Architectures

    A global media site serves the same static images to users worldwide from an S3 bucket in one Region. Which CloudFront behavior improves latency for this content?

    1. It assigns a static anycast IP address to the bucket
    2. It terminates client TLS at the S3 bucket
    3. It caches objects at edge locations closer to users Correct answer.
    4. It replicates the S3 bucket to every edge location

    Explanation

    CloudFront caches content at edge locations so that repeated requests are served from a location near the user rather than from the origin Region. This reduces round-trip time and origin load for static objects.

  129. Question 129Design High-Performing Architectures

    A company uploads large files from offices on several continents to a single S3 bucket in us-east-1. Users report slow uploads because of long-haul internet paths. Which S3 feature uses AWS edge locations to speed up these uploads?

    1. S3 multipart upload
    2. S3 Cross-Region Replication
    3. S3 Transfer Acceleration Correct answer.
    4. S3 Intelligent-Tiering

    Explanation

    S3 Transfer Acceleration routes uploads through AWS edge locations and the AWS global network to reach the bucket faster from distant locations. Cross-Region Replication and storage classes address durability or cost after upload, not long-haul upload speed.

  130. Question 130Design High-Performing Architectures

    A database server needs a low-cost EBS volume that provides burst capacity to 3,000 IOPS for occasional spikes, with baseline performance tied to volume size. Which volume type matches this requirement?

    1. Throughput Optimized HDD (st1)
    2. General Purpose SSD (gp2) Correct answer.
    3. General Purpose SSD (gp3)
    4. Provisioned IOPS SSD (io2)

    Explanation

    gp2 volumes earn burst credits and can burst to 3,000 IOPS, with baseline IOPS scaling at 3 IOPS per GiB of volume size. gp3 provides a baseline performance level independent of size, while io2 is provisioned and st1 is throughput-oriented.

  131. Question 131Design High-Performing Architectures

    An application uses ElastiCache and requires automatic failover, read replicas, and persistence for a session store. Which engine should be selected?

    1. DynamoDB Accelerator
    2. Redis Correct answer.
    3. Memcached
    4. Amazon RDS

    Explanation

    ElastiCache for Redis supports replication, automatic failover, and persistence options suitable for session stores. Memcached is a simpler in-memory cache without those data-protection features.

  132. Question 132Design High-Performing Architectures

    A Lambda function is CPU-bound and also makes many network calls. The team increases the configured memory from 512 MB to 2048 MB. What performance effect should they expect?

    1. The function is placed on a dedicated physical host
    2. The maximum execution timeout increases to 15 minutes
    3. CPU and network bandwidth allocated to the function increase Correct answer.
    4. Only the available disk space in /tmp increases

    Explanation

    Lambda allocates CPU power and network bandwidth in proportion to the configured memory. Increasing memory can therefore improve compute and network performance for CPU-bound or network-heavy functions.

  133. Question 133Design High-Performing Architectures

    A read-heavy web application uses a single RDS for MySQL instance and needs to offload reporting queries without changing the primary write endpoint. Which approach adds read capacity?

    1. Create one or more RDS read replicas Correct answer.
    2. Use an RDS Proxy in front of the primary
    3. Enable RDS Multi-AZ
    4. Increase the backup retention period

    Explanation

    RDS read replicas maintain a copy of the primary database and can serve read-only queries, offloading reporting traffic from the primary instance. Multi-AZ provides a standby for failover, not additional readable capacity.

  134. Question 134Design High-Performing Architectures

    An application must automatically switch DNS traffic from a primary endpoint to a standby endpoint when the primary fails health checks. Which Route 53 configuration should be used?

    1. Geolocation routing with a default record
    2. Failover routing with a health check on the primary record Correct answer.
    3. Weighted routing with equal weights
    4. Multivalue answer routing without health checks

    Explanation

    Failover routing uses a health check on the primary record and returns the standby record when the primary is unhealthy. This provides automated DNS-level failover between endpoints.

  135. Question 135Design High-Performing Architectures

    Users of a stateful web application must be routed to the same backend instance for the duration of their session behind an Application Load Balancer. Which setting should be enabled?

    1. Cross-zone load balancing
    2. HTTP/2 support
    3. Connection draining
    4. Target group stickiness Correct answer.

    Explanation

    Target group stickiness uses a cookie to bind a client session to the same target for the cookie lifetime. Cross-zone balancing and HTTP/2 affect distribution and protocol support, not session affinity.

  136. Question 136Design High-Performing Architectures

    A CloudFront distribution serves a popular API from a single origin, and cache misses from many edge locations cause repeated load on the origin. Which feature adds a centralized caching layer to reduce origin requests?

    1. Origin Shield Correct answer.
    2. CloudFront signed URLs
    3. Lambda@Edge
    4. Field-level encryption

    Explanation

    Origin Shield adds a centralized caching layer in front of the origin, reducing the number of cache-miss requests that reach the origin from many edge locations. It is designed to improve origin protection and cache hit ratio.

  137. Question 137Design High-Performing Architectures

    A client needs to download a single 5 GB S3 object as quickly as possible by issuing many parallel GET requests for different portions. Which S3 capability enables this?

    1. S3 Object Lock
    2. S3 Select
    3. S3 Inventory
    4. HTTP Range GET requests Correct answer.

    Explanation

    S3 supports HTTP Range GET requests, which retrieve a specified byte range of an object. Clients can issue multiple range requests in parallel to download different portions of a large object faster.

  138. Question 138Design High-Performing Architectures

    Multiple external partners call an API on API Gateway. The provider must give each partner a separate API key and enforce different request-rate limits. Which feature provides this?

    1. API Gateway caching per stage
    2. Resource policies with IP allow lists
    3. Usage plans with API keys and throttling limits Correct answer.
    4. Mutual TLS authentication

    Explanation

    Usage plans associate API keys with per-client throttling and quota limits, allowing different partners to receive different rate limits. API keys identify the client, while the usage plan defines the limits applied to that key.

  139. Question 139Design High-Performing Architectures

    A Lambda function shares an account's concurrency pool with many other functions. The team must ensure this function always has capacity for 100 concurrent invocations and cannot exceed that number. Which setting should be used?

    1. Provisioned concurrency set to 100
    2. Reserved concurrency set to 100 Correct answer.
    3. A Lambda alias with weighted routing
    4. An SQS event source with batch size 100

    Explanation

    Reserved concurrency sets aside a fixed amount of concurrency for a function and caps that function at the same value. This guarantees the function can use up to 100 concurrent invocations and prevents it from consuming more than that.

  140. Question 140Design High-Performing Architectures

    A new DynamoDB table will serve unpredictable traffic with long idle periods and sudden large spikes. The team wants to avoid capacity planning and throttling from insufficient provisioned capacity. Which capacity mode is appropriate?

    1. DynamoDB Accelerator (DAX)
    2. Provisioned capacity with reserved capacity
    3. Provisioned capacity with auto scaling
    4. On-demand capacity mode Correct answer.

    Explanation

    On-demand capacity mode automatically scales read and write capacity with actual traffic and requires no provisioned capacity planning. It is suitable for new or unpredictable workloads with sudden spikes.

  141. Question 141Design High-Performing Architectures

    A Lambda function opens a new database connection for each invocation and overwhelms an RDS for PostgreSQL instance with too many connections. Which service should be placed between Lambda and the database to pool and reuse connections?

    1. Amazon RDS Proxy Correct answer.
    2. Amazon ElastiCache
    3. Amazon RDS read replica
    4. AWS Global Accelerator

    Explanation

    RDS Proxy pools and reuses database connections, which reduces the connection overhead caused by many short-lived Lambda invocations. It sits between the application and the database and manages connections on the application's behalf.

  142. Question 142Design High-Performing Architectures

    An application sends messages to an SQS FIFO queue and needs high throughput while preserving order within each customer. Messages for different customers can be processed in parallel. What should the producer use?

    1. A single MessageGroupId for all messages
    2. Long polling with a 20-second wait time
    3. A distinct MessageGroupId per customer Correct answer.
    4. A standard queue instead of a FIFO queue

    Explanation

    SQS FIFO preserves order within each message group, and different message groups can be processed in parallel. Using a distinct MessageGroupId per customer maintains per-customer ordering while allowing concurrency across customers.

  143. Question 143Design High-Performing Architectures

    A Kinesis Data Firehose delivery stream must deliver records to S3 with low latency for near-real-time dashboards, even though this increases the number of S3 objects. Which setting should be adjusted?

    1. Increase the Firehose buffer size
    2. Reduce the Firehose buffer interval Correct answer.
    3. Enable S3 Object Lock
    4. Switch to S3 Standard-IA

    Explanation

    The Firehose buffer interval controls the maximum amount of time records are buffered before delivery. Reducing the interval lowers delivery latency, at the cost of producing more, smaller S3 objects.

  144. Question 144Design High-Performing Architectures

    A shared file system on EFS is accessed by thousands of EC2 instances with highly parallel workloads, and slightly higher per-operation latency is acceptable. Which performance mode should be selected?

    1. Elastic Throughput mode
    2. Max I/O performance mode Correct answer.
    3. General Purpose performance mode
    4. Provisioned Throughput mode

    Explanation

    EFS Max I/O performance mode is designed for highly parallel workloads and can scale to higher aggregate throughput and operations per second, accepting slightly higher per-operation latency. General Purpose mode prioritizes lower latency.

  145. Question 145Design High-Performing Architectures

    A company is migrating a Windows application that requires native SMB file shares, Active Directory integration, and Windows ACLs. Which AWS managed file system should be used?

    1. Amazon S3
    2. Amazon EFS
    3. Amazon FSx for Lustre
    4. Amazon FSx for Windows File Server Correct answer.

    Explanation

    Amazon FSx for Windows File Server provides native SMB file shares, Active Directory integration, and Windows ACL support. EFS is a Linux NFS file system, while S3 is object storage and Lustre is designed for high-performance computing.

  146. Question 146Design High-Performing Architectures

    A tightly coupled HPC workload runs on many EC2 instances and requires the lowest possible inter-node latency and high message-passing throughput. Which combination should be used?

    1. A spread placement group with Elastic Load Balancing
    2. A cluster placement group with Elastic Fabric Adapter Correct answer.
    3. An Auto Scaling group with a step scaling policy
    4. Multiple NAT gateways across Availability Zones

    Explanation

    A cluster placement group packs instances close together in a single Availability Zone to provide low-latency networking. Elastic Fabric Adapter provides high-throughput, low-latency networking for tightly coupled HPC and machine learning workloads.

  147. Question 147Design High-Performing Architectures

    A security team must transparently inspect all inbound and outbound traffic using third-party virtual firewall appliances that scale horizontally. Which load balancer type is designed for this use case?

    1. Network Load Balancer
    2. Classic Load Balancer
    3. Gateway Load Balancer Correct answer.
    4. Application Load Balancer

    Explanation

    Gateway Load Balancer is designed to distribute traffic to fleets of third-party virtual appliances such as firewalls and intrusion detection systems. It uses the GENEVE protocol to transparently route traffic through the appliance fleet.

  148. Question 148Design High-Performing Architectures

    A company has many VPCs and on-premises networks that must communicate through a central hub without creating a full mesh of VPC peering connections. Which service provides this hub-and-spoke connectivity?

    1. AWS PrivateLink
    2. VPC peering
    3. AWS Transit Gateway Correct answer.
    4. AWS Direct Connect gateway

    Explanation

    AWS Transit Gateway acts as a central hub that connects VPCs, VPNs, and Direct Connect gateways in a hub-and-spoke topology. This avoids the scaling complexity of a full mesh of VPC peering connections.

  149. Question 149Design High-Performing Architectures

    An application needs an active-active multi-Region DynamoDB deployment where each Region can accept writes and changes replicate to the other Regions. Which feature should be enabled?

    1. DynamoDB Streams with Lambda
    2. DynamoDB Accelerator (DAX)
    3. DynamoDB auto scaling
    4. DynamoDB global tables Correct answer.

    Explanation

    DynamoDB global tables provide a multi-Region, multi-active database in which each replica table can accept writes and changes are replicated to the other Regions. This supports active-active designs with low-latency local access.

  150. Question 150Design High-Performing Architectures

    A company wants to personalize HTML responses at CloudFront edge locations based on user cookies, without changing the origin. Which feature should be used?

    1. Lambda@Edge functions associated with cache behaviors Correct answer.
    2. S3 Transfer Acceleration
    3. CloudFront signed URLs
    4. AWS Global Accelerator

    Explanation

    Lambda@Edge runs custom code at CloudFront edge locations in response to viewer or origin events, allowing HTML personalization based on cookies at the edge. Signed URLs control access, while Transfer Acceleration and Global Accelerator address network transfer performance.

  151. Question 151Design Cost-Optimized Architectures

    A company runs steady-state EC2 workloads in eu-west-1 and wants a one-year commitment discount. During the year the workloads may change instance family, instance size, and even Region. Which purchasing option meets these requirements?

    1. EC2 Instance Savings Plan
    2. Compute Savings Plan Correct answer.
    3. Convertible Reserved Instance
    4. Standard Reserved Instance

    Explanation

    Compute Savings Plans apply to any EC2 usage regardless of instance family, size, tenancy, operating system, or Region, and they also cover Fargate and Lambda usage. Standard and EC2 Instance Savings Plans are tied to specific families or Regions, and Convertible Reserved Instances stay scoped to the Region in which they were purchased.

  152. Question 152Design Cost-Optimized Architectures

    A team stores regenerable thumbnail images in Amazon S3. The images can be lost if an Availability Zone fails, but millisecond access is required and storage cost must be as low as possible. Which storage class should be used?

    1. S3 One Zone-IA Correct answer.
    2. S3 Glacier Deep Archive
    3. S3 Standard-IA
    4. S3 Standard

    Explanation

    S3 One Zone-IA stores object data in a single Availability Zone, which lowers the storage price for data that can be recreated and does not need AZ-level resilience. It still provides millisecond access, unlike the Glacier classes.

  153. Question 153Design Cost-Optimized Architectures

    A nightly image-rendering job can be interrupted at any point and restarts from checkpoints. Which EC2 purchasing option minimizes the cost of running this job?

    1. Reserved Instances
    2. On-Demand Instances
    3. Dedicated Hosts
    4. Spot Instances Correct answer.

    Explanation

    Spot Instances use spare EC2 capacity at steep discounts and can be reclaimed with a short interruption notice, which suits workloads that checkpoint and resume. Reserved capacity and Dedicated Hosts are designed for predictable, interruption-free workloads.

  154. Question 154Design Cost-Optimized Architectures

    A finance team wants an automatic notification when the account's monthly spend is forecast to exceed $10,000. Which AWS service provides this?

    1. AWS Budgets Correct answer.
    2. AWS Trusted Advisor
    3. AWS Compute Optimizer
    4. AWS Cost Explorer

    Explanation

    AWS Budgets lets you define a spend or usage threshold and send notifications through Amazon SNS or email when actual or forecast values cross it. Cost Explorer visualizes and analyzes spend but does not deliver threshold alerts on its own.

  155. Question 155Design Cost-Optimized Architectures

    Which statement about Amazon EC2 data transfer charges is accurate?

    1. Data transfer from Amazon S3 to CloudFront is billed at internet egress rates.
    2. Data transfer between EC2 instances in the same Availability Zone over private IP is billed per GB.
    3. Inbound data transfer from the internet to EC2 is generally not charged. Correct answer.
    4. Data transfer between EC2 instances in different Availability Zones of the same Region is free.

    Explanation

    AWS generally does not charge for data transferred into its services from the internet, while egress and cross-AZ traffic are billed. Traffic between instances in the same Availability Zone over private IP addresses is free, and origin fetches from S3 to CloudFront are not charged as egress.

  156. Question 156Design Cost-Optimized Architectures

    A workload needs general-purpose SSD volumes with the ability to provision IOPS and throughput independently of volume size while keeping storage cost low. Which EBS volume type should be used?

    1. st1
    2. gp3 Correct answer.
    3. io2 Block Express
    4. gp2

    Explanation

    gp3 is a general-purpose SSD volume that lets you provision IOPS and throughput separately from capacity and is priced lower per GB than gp2. io2 targets high-performance transactional workloads at a higher price, and st1 is a throughput-oriented HDD.

  157. Question 157Design Cost-Optimized Architectures

    An internal application running on EC2 is used only between 08:00 and 18:00 on weekdays and must keep the same instance configuration. What is the most direct way to cut compute cost?

    1. Purchase three-year All Upfront Reserved Instances for the instances
    2. Enable detailed monitoring on the instances
    3. Stop the instances outside business hours on a schedule Correct answer.
    4. Migrate the instances to Dedicated Hosts

    Explanation

    Stopped instances do not incur compute charges, so scheduling shutdowns outside the usage window removes roughly two-thirds of the running hours without any application change. A commitment-based discount still pays for capacity during the idle hours.

  158. Question 158Design Cost-Optimized Architectures

    A company wants machine-learning-based rightsizing recommendations for its EC2 instances, EBS volumes, and Lambda functions. Which service should it use?

    1. AWS Compute Optimizer Correct answer.
    2. AWS Budgets
    3. AWS Trusted Advisor
    4. AWS Cost Explorer

    Explanation

    AWS Compute Optimizer analyzes utilization metrics with machine-learning models and returns rightsizing recommendations for resources such as EC2 instances, EBS volumes, Lambda functions, and ECS services on Fargate. It focuses on over- and under-provisioning rather than on spend thresholds or billing reports.

  159. Question 159Design Cost-Optimized Architectures

    A lifecycle rule must transition objects from S3 Standard to S3 Standard-IA. What is the minimum number of days an object must remain in S3 Standard before the transition can occur?

    1. 7 days
    2. 128 days
    3. 90 days
    4. 30 days Correct answer.

    Explanation

    S3 requires objects to be at least 30 days old before a lifecycle rule can transition them from S3 Standard into S3 Standard-IA or S3 One Zone-IA. Shorter intervals (for example 7 days) are rejected as invalid lifecycle configurations.

  160. Question 160Design Cost-Optimized Architectures

    A bucket holds millions of objects whose access patterns change frequently and cannot be predicted. The team wants automatic storage-class optimization with no retrieval fees for objects moved to the infrequent access tiers. Which storage class fits?

    1. S3 Intelligent-Tiering Correct answer.
    2. S3 One Zone-IA
    3. S3 Standard-IA with a 30-day lifecycle rule
    4. S3 Glacier Flexible Retrieval

    Explanation

    S3 Intelligent-Tiering monitors access patterns and moves objects between frequent and infrequent access tiers automatically, and the infrequent tiers incur no retrieval fees. It is designed exactly for unknown or shifting access patterns where a fixed lifecycle rule would guess wrong.

  161. Question 161Design Cost-Optimized Architectures

    A company has an AWS Organization with several member accounts and wants EC2 Reserved Instance discounts to apply across all Availability Zones in a Region and be shared with member accounts. Which configuration achieves this?

    1. Spot Instances combined with Savings Plans in each member account
    2. Zonal Reserved Instances that each member account purchases separately
    3. Convertible Reserved Instances with a zonal scope
    4. Regional Reserved Instances in the management account Correct answer.

    Explanation

    A regional Reserved Instance applies to matching usage anywhere in the Region and, with Reserved Instance sharing enabled in Organizations, its discount can be applied to usage in member accounts. Zonal Reserved Instances are tied to one Availability Zone and provide capacity reservation rather than regional flexibility.

  162. Question 162Design Cost-Optimized Architectures

    EC2 instances in a private subnet download several terabytes of data from Amazon S3 each month, and the NAT gateway data-processing charges are high. Which change removes those charges for S3 traffic?

    1. Attach an internet gateway directly to the private subnet.
    2. Enable S3 Transfer Acceleration on the bucket.
    3. Create an S3 gateway VPC endpoint Correct answer.
    4. Move the instances into a public subnet.

    Explanation

    A gateway VPC endpoint for S3 routes traffic to the service over the private AWS network and carries no hourly or data-processing charge, eliminating the NAT gateway hop. Only gateway endpoints for S3 and DynamoDB have this zero-cost property.

  163. Question 163Design Cost-Optimized Architectures

    A new serverless application has unpredictable, spiky traffic and the team does not want to provision read or write capacity in advance. Which Amazon DynamoDB configuration is most appropriate?

    1. DynamoDB Accelerator (DAX) in front of provisioned tables
    2. Provisioned mode with auto scaling
    3. Provisioned mode with reserved capacity
    4. On-demand capacity mode Correct answer.

    Explanation

    On-demand capacity mode charges per request and scales instantly, so an unpredictable workload never pays for idle provisioned capacity or throttles during a spike. Reserved capacity is a discount mechanism that only applies to provisioned throughput.

  164. Question 164Design Cost-Optimized Architectures

    Which statement about AWS Lambda pricing is correct?

    1. Provisioned concurrency is billed only when an invocation is processed.
    2. Lambda charges for the number of function versions stored.
    3. Lambda charges for data stored in the /tmp directory.
    4. Provisioned concurrency is billed while it is configured, even when idle. Correct answer.

    Explanation

    Provisioned concurrency keeps execution environments initialized and is billed for the concurrency and duration for which it is allocated, whether or not requests arrive. Invocation-based billing applies to on-demand execution, so provisioned capacity should be sized to a real baseline of traffic.

  165. Question 165Design Cost-Optimized Architectures

    An Amazon EFS file system stores project files. About 80 percent of the data has not been read for over 30 days. How can storage cost be reduced with minimal operational change?

    1. Use EFS lifecycle management to move files to Infrequent Access. Correct answer.
    2. Create a second file system in a different Region and replicate data to it.
    3. Move the files to instance store volumes on the application servers.
    4. Enable EFS provisioned throughput on the file system.

    Explanation

    EFS lifecycle management moves files that have not been accessed for a chosen period into the Infrequent Access storage class, which has a lower per-GB price, while the files remain readable through the same mount point. Serving a large cold subset from EFS IA typically cuts storage cost substantially with no application change.

  166. Question 166Design Cost-Optimized Architectures

    A development VPC has three Availability Zones, each with a private subnet and its own NAT gateway. High availability during an AZ failure is not required for this environment, and cost is the priority. What is the most cost-effective change?

    1. Add a fourth NAT gateway to distribute the load.
    2. Replace the three NAT gateways with a single NAT gateway. Correct answer.
    3. Replace the NAT gateways with an internet gateway on each private subnet.
    4. Move all private subnets into one Availability Zone and keep three NAT gateways.

    Explanation

    Each NAT gateway carries an hourly charge plus data-processing charges, so consolidating to one gateway in a non-production environment removes two-thirds of the fixed cost. The trade-off is that an AZ failure would interrupt egress, which this environment has stated it can accept.

  167. Question 167Design Cost-Optimized Architectures

    Compliance data must be retained for seven years and is almost never accessed. Retrieval within 12 hours is acceptable and storage cost is the top priority. Which S3 storage class should be used?

    1. S3 Glacier Flexible Retrieval
    2. S3 Glacier Deep Archive Correct answer.
    3. S3 Intelligent-Tiering
    4. S3 Standard-IA

    Explanation

    S3 Glacier Deep Archive has the lowest storage price of any S3 class and supports standard retrieval within 12 hours, which matches a long-retention, rarely read compliance archive. Glacier Flexible Retrieval costs more per GB, and the infrequent access classes cost far more.

  168. Question 168Design Cost-Optimized Architectures

    A company needs the most granular cost and usage data, broken down by resource ID and cost allocation tags, that can be queried with Amazon Athena. Which AWS feature provides this?

    1. AWS Trusted Advisor cost optimization checks
    2. AWS Cost and Usage Report delivered to Amazon S3 Correct answer.
    3. AWS Budgets cost reports
    4. AWS Cost Explorer monthly report grouped by service

    Explanation

    The AWS Cost and Usage Report publishes line-item billing data to S3 at resource and tag granularity and integrates with Athena for SQL analysis. Cost Explorer provides pre-aggregated views that are coarser than the CUR's line items.

  169. Question 169Design Cost-Optimized Architectures

    A research organization publishes a large dataset in Amazon S3. External users who have their own AWS accounts download it heavily, and the organization does not want to bear the data transfer cost. What should it configure?

    1. S3 Transfer Acceleration on the bucket
    2. Requester Pays on the bucket Correct answer.
    3. S3 Versioning with MFA delete
    4. S3 Cross-Region Replication to several Regions

    Explanation

    With Requester Pays enabled, the requester's AWS account is billed for requests and data transfer, while the bucket owner pays only for storage. Requesters must be authenticated and include the requester-pays header, which is why anonymous access cannot be combined with this setting.

  170. Question 170Design Cost-Optimized Architectures

    A company runs software licensed per physical socket and must bring its own license. Which EC2 tenancy option is both license-compliant and cost-effective?

    1. Shared tenancy with On-Demand Instances
    2. Dedicated Instances
    3. Spot Instances in multiple Regions
    4. Dedicated Hosts Correct answer.

    Explanation

    Dedicated Hosts give visibility and control over the physical sockets and cores, which per-socket or per-core bring-your-own-license terms require. Dedicated Instances isolate hardware but do not expose socket and core information needed for licensing.

  171. Question 171Design Cost-Optimized Architectures

    A fault-tolerant analytics job runs on Spot Instances and must keep running when individual Spot capacity pools are reclaimed. The job supports many instance types and can be restarted. Which configuration best maintains capacity at the lowest cost?

    1. On-Demand Instances in a spread placement group
    2. A mixed instances policy using capacity-optimized allocation Correct answer.
    3. A single instance type in one Availability Zone with the lowest-price allocation
    4. Reserved Instances spread across three Availability Zones

    Explanation

    Diversifying across many instance types and using the capacity-optimized allocation strategy places instances in the pools with the deepest available capacity, which lowers the frequency of simultaneous reclamation. Concentration in one pool with lowest-price allocation increases the chance that the whole fleet is interrupted at once.

  172. Question 172Design Cost-Optimized Architectures

    A company runs a steady 24/7 workload on m7g instances in us-east-1 and a spiky, unpredictable batch workload on AWS Fargate. Management wants to maximize savings with a one-year commitment while retaining flexibility. Which approach is most cost-effective?

    1. Purchase Convertible Reserved Instances for the Fargate workload.
    2. Standard Reserved Instances in each Availability Zone
    3. An EC2 Instance Savings Plan for the steady m7g usage only Correct answer.
    4. Purchase a Compute Savings Plan sized for the combined peak usage.

    Explanation

    EC2 Instance Savings Plans offer the deepest discount for a stable instance family in a fixed Region, and committing only to the steady baseline avoids paying for unused commitment on the spiky Fargate workload. Fargate has no reserved capacity product, so its variable usage is best left on demand or covered by an on-demand-sized Compute Savings Plan.

  173. Question 173Design Cost-Optimized Architectures

    A media company keeps 5 TB of archived footage in S3. Each object is accessed about once a quarter and must be retrievable in milliseconds when requested. Storage cost must be minimized. Which storage class is most appropriate?

    1. S3 Standard
    2. S3 Glacier Flexible Retrieval
    3. S3 Glacier Instant Retrieval Correct answer.
    4. S3 Standard-IA

    Explanation

    S3 Glacier Instant Retrieval is built for data accessed roughly once a quarter that still needs millisecond retrieval, and its storage price is lower than the Standard-IA class. Standard-IA costs more per GB and is intended for monthly access, while Glacier Flexible Retrieval cannot meet millisecond latency.

  174. Question 174Design Cost-Optimized Architectures

    A three-tier application in one Region transfers about 40 TB per month between its application tier in Availability Zone A and its database in Availability Zone B. The company wants to reduce inter-AZ data transfer charges while keeping the application highly available. Which change is most effective?

    1. Add a read replica in each AZ so reads stay in-AZ Correct answer.
    2. Enable S3 Transfer Acceleration for the database backups.
    3. Move the database to a single-AZ deployment and keep the application tier across three AZs.
    4. Replace the NAT gateways with VPC peering connections.

    Explanation

    Cross-AZ traffic is billed per GB in each direction, so serving reads from a replica in the same Availability Zone as each application instance removes most of the metered traffic while keeping the database resilient. Multi-AZ writes still cross AZs, but they are a small fraction of a read-heavy workload.

  175. Question 175Design Cost-Optimized Architectures

    A Lambda function is provisioned with 512 MB of memory and averages 6 seconds per invocation, and cost is dominated by GB-seconds. Benchmarking shows that increasing memory lowers duration. How should the team choose the most cost-effective memory setting?

    1. Always set the maximum memory because Lambda is billed only by invocation count.
    2. Benchmark candidate memory settings and choose the lowest GB-second cost Correct answer.
    3. Move the function to EC2 to remove Lambda duration charges.
    4. Keep 512 MB because memory has no effect on Lambda billing.

    Explanation

    Lambda duration cost equals allocated GB multiplied by execution time, so raising memory can reduce total cost when the function finishes proportionally faster. Testing candidate memory values and comparing GB-seconds per invocation finds the minimum rather than guessing.

  176. Question 176Design Cost-Optimized Architectures

    A company runs a mix of Amazon EC2 instances, AWS Fargate tasks, and AWS Lambda functions. It wants a single one- or three-year commitment that applies across all three compute services and can absorb changes in instance family and Region. Which purchase option meets these requirements?

    1. EC2 Instance Savings Plan
    2. Compute Savings Plan Correct answer.
    3. Convertible Reserved Instance
    4. Standard Reserved Instance

    Explanation

    Compute Savings Plans apply to EC2, Fargate, and Lambda usage in any Region and automatically cover changes in instance family, size, tenancy, and operating system. EC2 Instance Savings Plans and Reserved Instances are tied to a specific instance family and Region (or to EC2 only).

  177. Question 177Design Cost-Optimized Architectures

    A team stores thumbnail images in Amazon S3 that are regenerated from source photos whenever needed, so losing them in a single Availability Zone failure is acceptable. The thumbnails are read many times each day. Which storage class is the most cost-effective fit?

    1. S3 Standard-IA
    2. S3 One Zone-IA Correct answer.
    3. S3 Glacier Instant Retrieval
    4. S3 Standard

    Explanation

    S3 One Zone-IA stores data in a single Availability Zone and is priced below S3 Standard, which makes it appropriate for data that can be recreated and is not needed for durability across zones. It still offers millisecond access, unlike the archival classes.

  178. Question 178Design Cost-Optimized Architectures

    A 500 GB Amazon EBS volume is snapshotted every night, and only about 2 GB of the volume changes each night. How is the snapshot storage charged?

    1. Only the blocks that changed Correct answer.
    2. The full 500 GB for every snapshot taken
    3. The volume size each time the volume is modified
    4. The full 500 GB once; later snapshots are free

    Explanation

    EBS snapshots are incremental: each snapshot stores only the blocks that changed since the previous snapshot, plus any referenced blocks from earlier snapshots. Charges therefore reflect unique blocks rather than the full volume size on every run.

  179. Question 179Design Cost-Optimized Architectures

    Two Amazon EC2 instances in the same Availability Zone exchange several terabytes of data each month using their private IPv4 addresses. How is this traffic charged?

    1. At the cross-Availability-Zone data transfer rate
    2. At the NAT gateway data processing rate
    3. At the standard internet data transfer out rate
    4. No data transfer charge applies Correct answer.

    Explanation

    Data transferred between instances in the same Availability Zone over private IP addresses is not charged for data transfer. Cross-AZ, cross-Region, and internet egress traffic are the flows that incur charges.

  180. Question 180Design Cost-Optimized Architectures

    A global audience repeatedly downloads the same set of static images from an S3 bucket in us-east-1, and internet data transfer out is the largest line item on the bill. Which change most directly reduces that cost?

    1. Enable Requester Pays on the bucket
    2. Enable S3 Transfer Acceleration on the bucket
    3. Serve the objects through Amazon CloudFront Correct answer.
    4. Replicate the bucket to a second Region

    Explanation

    CloudFront caches objects at edge locations so repeated requests are served from the edge instead of the bucket, and edge data transfer rates are lower than direct S3 internet egress. The more repeat traffic, the larger the savings.

  181. Question 181Design Cost-Optimized Architectures

    For a given one-year or three-year term, which Reserved Instance payment option produces the largest discount relative to On-Demand pricing?

    1. Partial upfront
    2. Monthly payments with a cancellation option
    3. No upfront
    4. All upfront Correct answer.

    Explanation

    AWS prices Reserved Instances so that paying the entire term cost at purchase (All upfront) yields the deepest discount, followed by Partial upfront and then No upfront. The trade-off is a larger initial cash outlay.

  182. Question 182Design Cost-Optimized Architectures

    A finance team wants to be notified when the monthly cost of resources carrying a specific cost allocation tag is projected to exceed a threshold. Which AWS service provides this capability?

    1. AWS Trusted Advisor
    2. AWS Cost and Usage Report
    3. AWS Budgets Correct answer.
    4. AWS Cost Explorer

    Explanation

    AWS Budgets lets you define cost or usage budgets filtered by dimensions such as tags, and it sends alerts or triggers actions when actual or forecasted spend crosses a threshold. Cost Explorer and the Cost and Usage Report provide analysis data rather than threshold alerting.

  183. Question 183Design Cost-Optimized Architectures

    An EC2 instance is terminated, but the bill still shows storage charges for that workload. Which resource most likely explains the continued charge?

    1. The security group attached to the instance
    2. The EC2 key pair
    3. The instance store volumes
    4. An EBS volume with DeleteOnTermination disabled Correct answer.

    Explanation

    EBS volumes persist after instance termination unless the DeleteOnTermination attribute is enabled, so orphaned volumes keep incurring per-GB and IOPS charges. Security groups and key pairs are free, and instance store data is erased on termination.

  184. Question 184Design Cost-Optimized Architectures

    A fleet of EC2 instances uses 1 TB gp2 volumes with modest IOPS and throughput needs. The team wants to cut EBS cost without sacrificing general-purpose SSD performance. Which change should they make?

    1. Migrate the volumes to gp3 Correct answer.
    2. Enable EBS multi-attach on the volumes
    3. Enlarge each volume to gain more burst credits
    4. Migrate the volumes to io2

    Explanation

    gp3 is priced lower per GB than gp2 and includes a baseline of 3,000 IOPS and 125 MB/s regardless of volume size, with IOPS and throughput provisioned separately. For modest workloads it usually lowers cost without performance loss.

  185. Question 185Design Cost-Optimized Architectures

    Failed multipart uploads have left orphaned parts in an S3 bucket, and storage charges for that bucket keep climbing even though no new objects are being written. What is the most direct remedy?

    1. Abort incomplete multipart uploads with a lifecycle rule Correct answer.
    2. Transition the objects to S3 Standard-IA after 30 days
    3. Turn on bucket versioning and expire old versions after 30 days
    4. Enable S3 Transfer Acceleration on the bucket

    Explanation

    Parts from incomplete multipart uploads are billed as stored data and remain until they are explicitly aborted. A lifecycle rule with AbortIncompleteMultipartUpload removes them automatically after a set number of days.

  186. Question 186Design Cost-Optimized Architectures

    A reporting application runs heavy read queries against a production Amazon RDS database, and the team is considering moving to a much larger instance class to cope with the load. Which alternative is more cost-effective for offloading those reads?

    1. Convert the database to a Single-AZ deployment
    2. Extend automated backup retention to 35 days
    3. Enable RDS Multi-AZ for the production database
    4. Add a read replica and direct reports to it Correct answer.

    Explanation

    A read replica adds capacity only for read traffic and is typically cheaper than doubling the primary instance size. Multi-AZ standbys do not serve read traffic and therefore do not help a read-bound workload.

  187. Question 187Design Cost-Optimized Architectures

    A DynamoDB table uses provisioned capacity with steady, predictable read and write throughput that runs continuously all year. Which purchase option reduces the table's cost?

    1. Purchasing DynamoDB reserved capacity Correct answer.
    2. Amazon DynamoDB Accelerator (DAX)
    3. Switching the table to on-demand capacity mode
    4. Adding global tables in two additional Regions

    Explanation

    DynamoDB reserved capacity offers a discount on provisioned throughput units in exchange for a one- or three-year commitment, and it suits steady, predictable throughput. Changing capacity mode or adding replicas addresses other concerns.

  188. Question 188Design Cost-Optimized Architectures

    A batch container workload runs on Amazon ECS with the Fargate launch type. The work is stateless, can tolerate interruptions, and simply retries failed tasks. Which change reduces compute cost?

    1. Use Fargate Spot capacity providers Correct answer.
    2. Increase task CPU and memory allocation
    3. Enable ECS Exec on the running tasks
    4. Increase the task ephemeral storage size

    Explanation

    Fargate Spot uses spare capacity at a substantial discount, which is suitable for interruption-tolerant batch work. The task must be able to handle a two-minute interruption notice and retry.

  189. Question 189Design Cost-Optimized Architectures

    A company tags its resources by team, but the team dimension does not appear in Cost Explorer or in detailed billing reports. What must be done before costs can be grouped by team?

    1. Activate the cost allocation tags in the Billing console Correct answer.
    2. Create a separate AWS account for each team
    3. Enable Cost Explorer rightsizing recommendations
    4. Turn on AWS Config recording in every Region

    Explanation

    User-defined tags do not appear in cost reports until they are activated as cost allocation tags, and activation takes effect for usage after that point. Only activated tags become available as grouping dimensions.

  190. Question 190Design Cost-Optimized Architectures

    CloudWatch Logs charges are rising because log groups created months ago have never had a retention setting applied. What is the most direct cost control?

    1. Move the log groups to a different Region
    2. Increase the CloudWatch Logs agent buffer size
    3. Set a retention period on each log group Correct answer.
    4. Enable detailed monitoring on the EC2 instances

    Explanation

    By default, CloudWatch Logs keeps log events indefinitely, and storage is billed for all retained data. Setting a retention period deletes older events automatically and caps the ongoing storage charge.

  191. Question 191Design Cost-Optimized Architectures

    A company wants a three-year EC2 discount but expects to change instance families, sizes, and operating systems during that period. Which purchase option preserves that flexibility while still providing a discount?

    1. Standard Reserved Instances
    2. Dedicated Hosts
    3. On-Demand Instances
    4. Convertible Reserved Instances Correct answer.

    Explanation

    Convertible Reserved Instances can be exchanged for a new commitment of equal or greater value when attributes such as instance family or operating system change. Standard Reserved Instances are fixed for the term.

  192. Question 192Design Cost-Optimized Architectures

    A job needs a high-speed scratch area that is rebuilt from source data at the start of every run, and nothing on it must survive a stop or termination. Which storage choice is the most economical?

    1. Amazon EFS with provisioned throughput
    2. Instance store volumes Correct answer.
    3. Amazon FSx for Lustre
    4. Provisioned IOPS io2 volumes

    Explanation

    Instance store volumes are physically attached storage that is included with the instance, so there is no separate per-GB or per-IOPS charge. Because the data is temporary and regenerated each run, the lack of persistence is acceptable.

  193. Question 193Design Cost-Optimized Architectures

    CloudWatch metrics show that a 1 TB gp3 volume attached to an EC2 instance contains only about 200 GB of data. Which action reduces the storage charge while preserving the data?

    1. Detach the volume and reattach it as an io2 volume
    2. Create a smaller volume from a snapshot and attach it Correct answer.
    3. Take more frequent snapshots of the volume
    4. Enable EBS encryption on the volume

    Explanation

    EBS is billed on provisioned capacity, so unused space still costs money. Snapshotting the volume and restoring it into a smaller volume preserves the data while reducing the per-GB charge.

  194. Question 194Design Cost-Optimized Architectures

    An Amazon RDS for PostgreSQL database runs continuously with a fixed instance class and will remain in production for at least three years. Which pricing choice lowers cost the most?

    1. On-Demand instances with Multi-AZ enabled
    2. Copying automated snapshots to a second Region
    3. Spot Instances for the database host
    4. A three-year All Upfront Reserved Instance Correct answer.

    Explanation

    RDS Reserved Instances apply to a specific instance class in a Region and provide a substantial discount for a one- or three-year commitment, with All Upfront giving the largest reduction. Steady, long-lived workloads are the ideal fit.

  195. Question 195Design Cost-Optimized Architectures

    A log-processing cluster writes very large files sequentially to Amazon EBS and almost never rereads them; access, when it happens, is also sequential. Cost per GB is the dominant requirement. Which volume type fits best?

    1. io2
    2. sc1 Correct answer.
    3. st1
    4. gp3

    Explanation

    sc1 (Cold HDD) has the lowest cost per GB EBS volume type and is designed for infrequently accessed, large, sequential workloads. It trades performance for price, which is acceptable here.

  196. Question 196Design Cost-Optimized Architectures

    A bucket contains 40 million objects averaging 60 KB, each read a few times per year. A lifecycle policy transitions them to S3 Standard-IA after 30 days. Why might this transition fail to reduce cost as expected?

    1. Standard-IA cannot store objects smaller than 1 MB
    2. Standard-IA enforces a 90-day minimum storage duration
    3. Objects under 128 KB are billed as 128 KB in Standard-IA Correct answer.
    4. Lifecycle rules cannot transition objects to Standard-IA

    Explanation

    S3 Standard-IA (and other IA and archive classes) bills objects smaller than 128 KB as if they were 128 KB, so small objects see little or no saving. With 60 KB objects the effective price per object is more than double the nominal per-GB rate.

  197. Question 197Design Cost-Optimized Architectures

    A company wants to purchase a Compute Savings Plan but does not know what hourly commitment to make. Which capability helps identify a commitment that covers a chosen portion of its steady-state usage?

    1. Cost Explorer Savings Plans recommendations Correct answer.
    2. CloudWatch dashboards showing CPU utilization
    3. Trusted Advisor cost optimization checks
    4. AWS Budgets with a fixed monthly spending limit

    Explanation

    Cost Explorer generates Savings Plans purchase recommendations based on historical usage, showing the commitment level, estimated savings, and coverage for different lookback periods. That lets the team pick a commitment matching its risk tolerance.

  198. Question 198Design Cost-Optimized Architectures

    More than 30 percent of the total cost of an Amazon Aurora PostgreSQL cluster comes from I/O charges, driven by a read-heavy transactional workload. Which change is designed to reduce that cost?

    1. Enable Aurora I/O-Optimized Correct answer.
    2. Add a second reader instance
    3. Increase the instance class of the writer
    4. Switch the cluster to Aurora Standard

    Explanation

    Aurora I/O-Optimized removes per-I/O charges in exchange for a higher storage and instance price, and it becomes cheaper when I/O spend exceeds roughly 25 percent of total Aurora cost. This cluster is above that threshold.

  199. Question 199Design Cost-Optimized Architectures

    A company continues to run Amazon RDS for MySQL 5.7 well past the end of standard support while it postpones a major version upgrade. What cost consequence should it expect?

    1. Automated backups are billed at double the standard rate
    2. Standard support continues with no pricing change
    3. RDS Extended Support charges apply per vCPU-hour Correct answer.
    4. The database instance is stopped automatically

    Explanation

    For engines that have passed end of standard support, RDS charges RDS Extended Support on a per-vCPU-hour basis for continued operation on the older major version. Upgrading to a supported version avoids that surcharge.

  200. Question 200Design Cost-Optimized Architectures

    A team keeps monthly Amazon EBS snapshots for five years for audit purposes. Restores are rare and can be completed within 72 hours. How can the snapshot storage cost be reduced the most?

    1. Copy the snapshots to a second Region
    2. Move the snapshots to the EBS snapshot archive tier Correct answer.
    3. Convert the snapshots into AMIs
    4. Enable fast snapshot restore for the snapshots

    Explanation

    The EBS snapshot archive tier is priced well below standard snapshot storage for snapshots that are kept long term and rarely restored. Restoring an archived snapshot can take up to 72 hours, which matches this requirement.

AWS Solutions Architect flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

IAM explicit deny
An explicit Deny in any applicable policy overrides every Allow, including administrator access and resource-based grants; policy evaluation returns deny immediately.
Service control policy scope
SCPs cap the maximum permissions available to member accounts in AWS Organizations; they never grant permissions and do not affect the management account.
Security group vs NACL
Security groups are stateful, allow-only filters attached to elastic network interfaces; network ACLs are stateless, support allow and deny rules, and are evaluated per subnet in rule-number order.
KMS key policy requirement
A KMS key policy must delegate to IAM before identity-based policies can use the key; otherwise only the principals named directly in the key policy have access.
S3 SSE-KMS audit trail
SSE-KMS logs key usage to CloudTrail and supports separate permissions per key, while SSE-S3 uses AWS-managed keys with no per-key access control or detailed audit trail.
Gateway vs interface endpoint
Gateway endpoints for S3 and DynamoDB use route table entries and carry no hourly charge; interface endpoints create ENIs with security groups and bill hourly plus per GB.
Secrets Manager rotation
Secrets Manager rotates credentials automatically on a schedule by invoking a Lambda rotation function; SSM Parameter Store SecureString only stores the secret and requires custom rotation code.
Cognito identity pool role
Identity pools exchange authenticated or guest tokens for temporary AWS credentials through STS, scoped by an IAM role and its attached policy.
RTO versus RPO
RTO is the maximum tolerable downtime before service resumes; RPO is the maximum tolerable data loss measured backwards from the failure. Both drive disaster recovery design.
Warm standby strategy
A scaled-down but fully functional copy of production runs in another Region and is scaled up on failover; it costs more than pilot light but delivers a lower RTO.
RDS Multi-AZ failover
RDS Multi-AZ keeps a synchronous standby in another Availability Zone and fails over automatically, typically in 60 to 120 seconds, using the same DNS endpoint.
SQS visibility timeout
A received message stays hidden for the visibility timeout; if the consumer does not delete it in time, the message becomes visible again and is redelivered.
SQS FIFO deduplication
FIFO queues deliver exactly-once processing within a five-minute deduplication interval using a message deduplication ID, and preserve strict ordering within each message group.
Route 53 failover routing
Failover routing sends traffic to a primary record while its health check passes and diverts to a secondary record once that check reports unhealthy.
DynamoDB global tables
Global tables replicate a DynamoDB table into multiple Regions with last-writer-wins conflict resolution, enabling multi-Region active-active reads and writes.
S3 One Zone-IA risk
One Zone-IA stores objects in a single Availability Zone, so an AZ failure destroys the data; it suits reproducible data that does not need resilience.
Aurora storage replication
Aurora keeps six copies of data across three Availability Zones, supports up to 15 read replicas, and exposes a reader endpoint that load-balances incoming connections.
gp3 versus gp2
gp3 delivers a 3,000 IOPS and 125 MB/s baseline independent of volume size, while gp2 performance scales with capacity at 3 IOPS per GiB.
Cluster placement group
Cluster placement groups pack instances close together inside a single Availability Zone for low latency and high throughput, typical of HPC and tightly coupled workloads.
ALB versus NLB
ALB routes Layer 7 requests by host, path, or header and terminates TLS; NLB handles Layer 4 at extreme throughput, preserves source IP, and provides static IP addresses.
CloudFront origin groups
Origin groups pair a primary and a secondary origin so CloudFront fails over automatically when the primary returns configured error codes or times out.
CloudFront Functions vs Lambda@Edge
CloudFront Functions run lightweight JavaScript at viewer events in sub-millisecond time; Lambda@Edge handles heavier logic at viewer and origin request events in Node.js or Python.
S3 multipart upload
Multipart upload splits an object into parts transferred in parallel, improving speed and allowing resume; it is recommended above 100 MB and required above 5 GB.
DynamoDB hot partitions
Low-cardinality partition keys concentrate traffic on a few partitions; high-cardinality keys combined with on-demand mode or adaptive capacity spread the load.
Compute Savings Plans
Compute Savings Plans commit to hourly spend for one or three years and apply automatically to EC2, Fargate, and Lambda across families and Regions, saving up to 66 percent.
EC2 Instance Savings Plans
These plans lock to one instance family in one Region in exchange for deeper discounts of up to 72 percent, trading flexibility for a lower hourly rate.
Spot Instance interruption
Spot Instances offer up to 90 percent off but can be reclaimed with a two-minute warning, so they suit stateless, checkpointed, or fault-tolerant workloads.
S3 Intelligent-Tiering
Intelligent-Tiering moves objects between access tiers automatically based on observed access patterns, charging a small monitoring fee instead of retrieval fees for tier changes.
Glacier minimum storage duration
S3 Glacier classes bill a minimum storage period, 90 days for Glacier tiers and 180 days for Deep Archive, even if objects are deleted earlier.
NAT Gateway cost driver
A NAT Gateway bills an hourly charge plus per-GB data processing; routing S3 and DynamoDB traffic through a free gateway endpoint avoids those processing charges entirely.
Data transfer into AWS
Inbound data transfer to AWS is generally free, while outbound transfer to the internet is billed per GB, so egress volume typically dominates cost design.
EBS snapshot incrementality
EBS snapshots are incremental after the first full copy and billed only for changed blocks; deleting a snapshot retains blocks still referenced by later snapshots.

AWS Solutions Architect study guide

Design Secure Architectures

Core concepts

Security questions on this exam are usually about who can do what, from where, and with what encryption. Know the IAM evaluation chain: an explicit deny always wins, then an explicit allow, otherwise the request is implicitly denied. Enumerate the policy types and what each can and cannot do: identity-based policies, resource-based policies, permissions boundaries (they cap, never grant), SCPs (they cap an entire account or OU, never grant), and session policies.

On the network side, be precise about statefulness. Security groups are stateful, allow-only, and reference other security groups rather than CIDRs where possible. NACLs are stateless and ordered, so return traffic needs ephemeral ports (1024-65535) opened. Know which service is the right detector: GuardDuty for threat detection, Inspector for vulnerability/CVE scanning, Macie for PII discovery in S3, Config for compliance drift, CloudTrail for API audit, Security Hub for aggregation.

Encryption: SSE-S3 versus SSE-KMS versus client-side, TLS in transit, and the fact that a KMS key policy is the root of authority for a CMK. For edge protection, know Shield (DDoS) versus WAF (Layer 7 rules) versus ACM.

Traps

  • Cross-account access almost always needs both sides: a role/policy in the trusted account and a resource or key policy in the owning account.
  • Questions offering "create an IAM user with access keys on the instance" are distractors; the answer is an instance profile role.
  • S3 Block Public Access and Object Ownership can override a bucket policy that looks correct.

Worked example

An EC2 app must read objects from a bucket in another account, encrypted with a customer managed key. The complete answer is: an instance profile role on EC2, a bucket policy granting that role ARN s3:GetObject, a KMS key policy granting the role kms:Decrypt, and no long-term credentials anywhere.

Design Resilient Architectures

Core concepts

Resilience questions are framed by RPO (data loss tolerated) and RTO (downtime tolerated). Read those numbers first; they eliminate most options. Then reason in layers.

Compute and networking: spread across at least two AZs using an Auto Scaling group with subnets in each AZ, behind an ELB that health-checks and drains connections. A single NAT gateway is a classic single point of failure; one per AZ is the resilient answer. Elastic IPs and instance store are anti-patterns for durability (instance store is ephemeral).

Data: RDS Multi-AZ is for failover (synchronous standby, no read offload); read replicas are for read scaling and are asynchronous. Aurora offers replicas, backtrack, and Global Database for cross-region. S3 gives durability via redundancy plus versioning and cross-region replication. DynamoDB global tables give multi-region active-active.

Decoupling: SQS between tiers absorbs spikes; set a dead-letter queue for poison messages, and remember visibility timeout versus consumer processing time. Step Functions and EventBridge coordinate long workflows.

DR strategies from cheapest/slowest to most expensive/fastest: backup and restore, pilot light, warm standby, multi-site active-active.

Traps

  • "Highly available" does not mean "Multi-AZ RDS increases read throughput."
  • Deploying two instances in the same subnet or one AZ is not HA.
  • SQS FIFO has lower throughput and does not support every feature standard queues do.

Worked example

A workload requires RTO of 15 minutes and RPO near zero across regions. Backups alone fail the RTO; multi-site active-active is overkill. Warm standby - a scaled-down but fully functional copy in a second region, continuous data replication, and Route 53 failover routing - fits.

Design High-Performing Architectures

Core concepts

Performance questions reward picking the right tool for the bottleneck and knowing the scaling lever each service offers.

Compute: match instance families to workload (compute, memory, storage, accelerated), consider Graviton for price-performance, and use placement groups - cluster for low-latency HPC (single AZ), spread for small numbers of critical instances, partition for distributed systems like HDFS and Kafka. Enhanced networking and EFA matter for tightly coupled workloads.

Storage: gp3 decouples IOPS from capacity (unlike gp2), io2 Block Express for high-IOPS databases, st1 for throughput-heavy sequential access, instance store for scratch. S3 performance scales with request parallelism: multipart upload for large objects, byte-range fetches, and prefix distribution; Transfer Acceleration for distant uploaders.

Databases: Aurora reader endpoints and read replicas for read-heavy traffic, ElastiCache/DAX for hot-read caching, and DynamoDB partition-key design to avoid hot partitions. Adaptive capacity helps but does not fix a low-cardinality key.

Decoupling and edge: CloudFront for cacheable content at the edge, Global Accelerator for TCP/UDP latency over the AWS backbone, ALB for HTTP path/host routing, NLB for extreme throughput, static IPs, or non-HTTP protocols.

Traps

  • Choosing a bigger instance when the real fix is caching, indexing, or partitioning.
  • Assuming DynamoDB provisioned throughput auto-scales instantly under a hot key.
  • Using a Network Load Balancer for path-based routing (that is ALB).

Worked example

A DynamoDB table throttles under load even with high provisioned capacity; the key is a timestamp. Reads for "today" all hit one partition. Fix the key design (add a high-cardinality suffix or a write-sharding component) rather than adding capacity.

Design Cost-Optimized Architectures

Core concepts

Cost questions test whether you can identify the shape of the workload and the hidden line items.

Purchasing: On-Demand for spiky or short-lived work, Savings Plans and Reserved Instances for steady-state baselines (Compute Savings Plans are the most flexible across families and regions), Spot for fault-tolerant batch and stateless workers with checkpointing. A mixed answer is usually correct: cover the 24/7 baseline with a commitment, run the burst on On-Demand or Spot.

Storage lifecycle: S3 Standard to Standard-IA to Glacier Instant Retrieval, Flexible Retrieval, Deep Archive, driven by access pattern and retrieval tolerance; Intelligent-Tiering when patterns are unknown. Lifecycle rules expire noncurrent versions and incomplete multipart uploads. gp2 volumes can move to gp3 for roughly 20% savings at equal-or-better baseline performance, and unattached EBS volumes and idle Elastic IPs are pure waste.

Data transfer and network: the cheapest byte is the one that never crosses a boundary. Cross-AZ traffic is billed in both directions, NAT Gateway charges per GB processed plus hourly, and cross-region transfer is expensive. CloudFront and VPC gateway endpoints for S3/DynamoDB often cut NAT costs substantially. Keep chatty tiers in one AZ when resilience allows.

Visibility: Cost Explorer, Budgets with alerts, Cost Allocation Tags, and Trusted Advisor checks are the governance tools; Compute Optimizer and S3 Storage Lens suggest right-sizing.

Traps

  • Reserved capacity for a workload that runs four hours a week.
  • Spot for stateful, non-interruptible workloads.
  • Forgetting NAT Gateway data processing charges when sizing a VPC design.

Worked example

A steady 24/7 web tier with unpredictable nightly batch: buy Compute Savings Plans for the always-on baseline, run the batch on Spot with checkpointing, and put static assets behind CloudFront to reduce both egress and origin load.

AWS Solutions Architect cheat sheet

Quick reference

Design Secure Architectures

  • Policy evaluation order: explicit Deny > Organizations SCP > resource-based policy > identity-based policy > permissions boundary > session policy; anything not explicitly allowed is implicitly denied.

  • Cross-account access needs permission on both sides: an identity-based policy in the caller account plus a resource-based policy or role trust policy in the owning account.

  • Security groups are stateful and allow-only and attach to ENIs; network ACLs are stateless, support both allow and deny, and are evaluated by rule number at the subnet boundary.

  • A KMS key policy is the primary access control for a customer managed key; IAM policies grant nothing unless the key policy delegates to IAM.

  • S3 Block Public Access set at the account level overrides bucket policies and object ACLs, so a permissive bucket policy alone will not open access.

  • Gateway VPC endpoints (S3, DynamoDB) work through route table entries and have no security group; interface endpoints (PrivateLink) create ENIs with security groups.

  • Secrets Manager performs scheduled automatic rotation via a Lambda rotation function; Parameter Store SecureString stores secrets cheaply but rotation must be implemented manually.

  • Cognito user pools authenticate users and issue OIDC tokens; identity pools exchange those tokens for temporary AWS credentials through STS scoped by an IAM role.

  • AWS Shield Standard is automatic and free for all accounts; Shield Advanced adds DDoS cost protection, SRT access, and Layer 7 mitigation for a subscription fee.

Quick reference

Design Resilient Architectures

  • RTO is maximum tolerable downtime, RPO is maximum tolerable data loss; backup-and-restore has the highest RTO/RPO and multi-site active-active the lowest.

  • DR strategies ordered by rising cost and falling RTO: backup and restore, pilot light, warm standby, multi-site active-active.

  • RDS Multi-AZ keeps a synchronous standby with automatic failover in roughly 60-120 seconds on the same DNS name; read replicas are asynchronous and need manual promotion.

  • Aurora maintains six copies of data across three AZs and fails over to a read replica, typically in under 30 seconds.

  • SQS visibility timeout should be at least six times the consumer or Lambda timeout so messages are not redelivered while still being processed.

  • Standard SQS queues give at-least-once delivery with best-effort ordering; FIFO queues give exactly-once processing per message group using a deduplication ID.

  • An Auto Scaling group spanning multiple subnets in different AZs should use ELB health checks so instances failing application checks are replaced, not just failed EC2 status checks.

  • Route 53 failover routing requires a health check on the primary record and sends traffic to the secondary record when that check reports unhealthy.

  • S3 One Zone-IA stores data in a single AZ and loses it if that AZ fails; use versioning plus Cross-Region Replication when regional loss must be survivable.

Quick reference

Design High-Performing Architectures

  • EBS gp3 gives a 3,000 IOPS and 125 MB/s baseline regardless of volume size; gp2 scales at 3 IOPS per GiB, so small gp2 volumes are IOPS-starved.

  • Use instance store for the highest IOPS at the cost of ephemerality; use io2/io1 EBS when high IOPS must persist across stop/start and AZ moves.

  • Placement group rules: cluster for low-latency HPC within one AZ, spread for a maximum of seven instances per AZ per group, partition for HDFS and Kafka.

  • Aurora provides up to 15 read replicas and a reader endpoint that load-balances connections; the cluster endpoint always points at the writer.

  • ALB is Layer 7 and routes by host, path, or header; NLB is Layer 4, preserves source IP, offers static IPs, and sustains millions of requests per second.

  • CloudFront origin groups pair primary and secondary origins so the distribution fails over on specified HTTP status codes or origin timeouts.

  • CloudFront Functions run sub-millisecond JavaScript at viewer request/response; Lambda@Edge handles heavier logic at viewer and origin events in Node.js or Python.

  • S3 multipart upload parallelizes parts for speed and resumability; it is recommended above 100 MB and required for objects larger than 5 GB.

  • DynamoDB throughput concentrates on partitions with hot keys; use high-cardinality partition keys with on-demand mode or adaptive capacity to spread load.

Quick reference

Design Cost-Optimized Architectures

  • Compute Savings Plans apply across EC2, Fargate, and Lambda in any family or Region for up to 66% savings; EC2 Instance Savings Plans reach about 72% but lock to family and Region.

  • Spot Instances discount up to 90% with a two-minute interruption notice, so use them for stateless, checkpointed, or fault-tolerant batch work.

  • S3 Intelligent-Tiering charges a small monitoring fee and moves objects between access tiers automatically, avoiding retrieval fees that lifecycle transitions incur.

  • Minimum storage durations apply: S3 Standard-IA bills 30 days, Glacier Flexible and Instant Retrieval 90 days, and Deep Archive 180 days even if objects are deleted earlier.

  • Data transfer into AWS is generally free while egress to the internet is billed per GB, so architectures that keep traffic inside a Region or behind CloudFront cost less.

  • A NAT Gateway bills hourly plus per GB processed; routing S3 and DynamoDB traffic through a free gateway VPC endpoint removes that data processing charge.

  • EBS snapshots are incremental after the first copy and billed only for changed blocks; retain the chain carefully because deleting a parent leaves dependent blocks billed.

  • Reserved Instances are Standard (fixed family) or Convertible (exchangeable), in 1- or 3-year terms; only zonal RIs reserve capacity in a specific AZ.

  • Use cost allocation tags with the Cost and Usage Report for granular attribution, and AWS Budgets for threshold alerts rather than relying on billing alarms alone.

AWS Solutions Architect key terms

Identity-Based Policy

JSON permission document attached to an IAM user, group, or role that lists which actions are allowed or denied on which resources. It defines what a principal may do, not who may assume it.

Related: Resource-Based Policy, Service Control Policy, AWS STS AssumeRole
Resource-Based Policy

Permission document attached directly to an AWS resource, such as an S3 bucket or SQS queue, naming which principals may act on it and under what conditions. Commonly used to grant cross-account access.

Related: Identity-Based Policy, AWS STS AssumeRole
Security Group

Stateful virtual firewall that controls inbound and outbound traffic for elastic network interfaces and the resources using them. Only allow rules exist, and return traffic for permitted connections is automatically allowed.

Related: Network ACL, VPC Endpoint
Network ACL

Stateless, numbered allow and deny filter applied at the subnet boundary. Rules are evaluated in order and apply separately to each direction, so return traffic requires its own rule.

Related: Security Group, VPC Endpoint
AWS KMS

Managed service that creates and stores encryption keys and performs cryptographic operations through API calls. Keys never leave the service, and access is governed by key policies together with IAM.

Related: AWS Secrets Manager, Resource-Based Policy
VPC Endpoint

Private connection from a VPC to a supported AWS service without traversing the public internet. Gateway endpoints serve S3 and DynamoDB; interface endpoints use elastic network interfaces with private IP addresses.

Related: Security Group, Network ACL
Service Control Policy

Organization-level policy in AWS Organizations that sets the maximum permissions available to member accounts. It filters permissions but never grants them, so an action must also be allowed by IAM.

Related: Identity-Based Policy, AWS STS AssumeRole
AWS WAF

Web application firewall for CloudFront, Application Load Balancer, and API Gateway that inspects HTTP requests and allows, blocks, or counts them using rules, rate limits, and managed rule groups.

Related: AWS Shield Advanced, Amazon CloudFront, Application Load Balancer
AWS Shield Advanced

Paid DDoS protection offering enhanced detection, near-real-time attack visibility, access to the AWS response team, and cost protection for scaling charges incurred during an attack.

Related: AWS WAF, Amazon CloudFront
Amazon Cognito

Identity service providing user directories for sign-up and sign-in (user pools) plus temporary AWS credentials for authenticated app users (identity pools), commonly used by mobile and web applications.

Related: AWS STS AssumeRole, Application Load Balancer
AWS STS AssumeRole

API call that returns temporary, expiring credentials for an IAM role. It underpins cross-account access, identity federation, and replacing long-term access keys on workloads.

Related: Identity-Based Policy, Service Control Policy, Amazon Cognito
AWS Secrets Manager

Store for database credentials and API keys with built-in rotation through Lambda, fine-grained access control, and native integration with services such as RDS and Redshift.

Related: AWS KMS
Multi-AZ Deployment

Running resources across two or more Availability Zones in a region so that a single data center or zone failure does not take the workload offline. Managed databases use a synchronous standby in another zone.

Related: Read Replica, Recovery Time Objective, Route 53 Failover Routing
Read Replica

Asynchronous copy of a database that serves read traffic and can be promoted to primary. It reduces load on the writer and offers a recovery target, but may lag behind the source.

Related: Multi-AZ Deployment, Recovery Point Objective
Route 53 Failover Routing

DNS policy that directs traffic to a primary record while health checks pass and automatically switches to a secondary record when the primary is found unhealthy.

Related: Multi-AZ Deployment, Route 53 Latency-Based Routing
Auto Scaling Group

Collection of EC2 instances managed as one unit with minimum, maximum, and desired capacity. It replaces unhealthy instances and adds or removes capacity in response to demand.

Related: Multi-AZ Deployment, Right Sizing
Recovery Time Objective

Maximum acceptable elapsed time to restore a workload after a disruption. It drives architecture choices such as warm standby versus backup and restore.

Related: Recovery Point Objective, Pilot Light
Recovery Point Objective

Maximum acceptable amount of data loss, expressed as time before the failure. It determines how frequently backups run and how replication is configured.

Related: Recovery Time Objective, S3 Cross-Region Replication
Pilot Light

Disaster recovery pattern in which a minimal core of replicated data and a few core servers runs in the recovery region and is scaled out at failover. The cheapest of the active recovery patterns.

Related: Recovery Time Objective, Recovery Point Objective, AWS Backup
S3 Cross-Region Replication

Automatic asynchronous copying of objects to a bucket in a different region, used for compliance, lower-latency access, and regional failure resilience. Versioning is required on both buckets.

Related: Recovery Point Objective, DynamoDB Global Tables
DynamoDB Global Tables

Multi-region, multi-active replication for DynamoDB tables in which every replica accepts writes and converges. It provides low-latency local reads and regional failover for NoSQL workloads.

Related: Multi-AZ Deployment, S3 Cross-Region Replication
AWS Backup

Centralized, policy-driven service that schedules, retains, and restores backups across many AWS services, including cross-region and cross-account copies for compliance and recovery.

Related: Pilot Light, Recovery Point Objective
Application Load Balancer

Layer 7 load balancer that routes HTTP and HTTPS requests by path, host, header, or query string, supports WebSocket, and integrates with AWS WAF and Amazon Cognito authentication.

Related: Network Load Balancer, AWS WAF, Amazon CloudFront
Network Load Balancer

Layer 4 load balancer built for extreme throughput and millions of requests per second with very low latency. It preserves source IP addresses and supports static Elastic IP addresses per zone.

Related: Application Load Balancer, AWS Global Accelerator
Amazon CloudFront

Global content delivery network that caches responses at edge locations to cut latency and reduce origin load, with TLS termination, signed URLs, and WAF integration.

Related: AWS Global Accelerator, AWS WAF
Amazon ElastiCache

Managed in-memory caching service using Redis or Memcached that serves repeated reads from memory, offloads databases, and can hold session state for stateless application tiers.

Related: DynamoDB Accelerator, Read Replica
DynamoDB Accelerator

In-memory write-through cache for DynamoDB that reduces read latency from milliseconds to microseconds for read-heavy tables, requiring only an endpoint change rather than application rewrites.

Related: Amazon ElastiCache
Cluster Placement Group

Logical grouping that packs instances close together within a single Availability Zone on high-bandwidth, low-latency networking, suited to tightly coupled high performance computing workloads.

Related: Network Load Balancer
AWS Global Accelerator

Service that uses AWS edge locations and anycast IP addresses to carry TCP and UDP traffic over the AWS backbone to the best endpoint, improving latency and failover speed.

Related: Amazon CloudFront, Route 53 Latency-Based Routing, Network Load Balancer
Route 53 Latency-Based Routing

DNS policy that answers with the record for the region measured to have the lowest latency for the requester, often combined with health checks to avoid slow or failed regions.

Related: Route 53 Failover Routing, AWS Global Accelerator
Savings Plans

Commitment to a consistent hourly amount of compute usage for a one- or three-year term in exchange for lower rates. Compute Savings Plans apply across instance families, sizes, and regions automatically.

Related: Reserved Instances, Right Sizing
Reserved Instances

Discounted capacity reservation for services such as EC2, RDS, and ElastiCache over a one- or three-year term. Standard reservations give the deepest discount but the least flexibility to change.

Related: Savings Plans, Spot Instances
Spot Instances

Spare EC2 capacity offered at steep discounts that can be reclaimed with a two-minute interruption notice. Best suited to fault-tolerant, stateless, and batch workloads rather than persistent stateful servers.

Related: Reserved Instances, Auto Scaling Group
S3 Intelligent-Tiering

Storage class that automatically moves objects between frequent-access and infrequent-access tiers based on observed usage, charging a small monitoring fee instead of retrieval charges.

Related: S3 Lifecycle Policy, Right Sizing
S3 Lifecycle Policy

Set of rules that transitions objects between storage classes or expires them after defined periods, automating retention compliance and removing the cost of storing cold data in hot tiers.

Related: S3 Intelligent-Tiering
Right Sizing

Practice of matching instance types, storage, and database capacity to measured demand, typically using utilization metrics and recommendations, so that over-provisioned spend is eliminated.

Related: AWS Cost Explorer, Savings Plans, Auto Scaling Group
AWS Cost Explorer

Console and API tool for visualizing, filtering, and forecasting spend by service, account, or tag, and for reviewing Savings Plans and Reserved Instance purchase recommendations.

Related: Savings Plans, Right Sizing

AWS Solutions Architect study plan

8 weeks · About 6 hours per week

Week 1

Design Secure Architectures

  • Map every AWS identity type (IAM user, group, role, federated identity, resource-based policy) and write the evaluation logic for a single request in your own words

  • Build three IAM policies from scratch: a least-privilege S3 bucket policy, a role trust policy for cross-account access, and a permissions boundary that caps a developer role

  • Differentiate service control policies from identity policies by sketching an Organizations tree with an OU-level deny that overrides an account admin

  • Compare KMS key policies, grants, and IAM policy interaction, then decide which one governs a given decrypt call

  • Enable and interpret CloudTrail, GuardDuty, and Config findings on one account to explain the detective-control layer

Your checkpoint

A one-page identity decision matrix (identity type -> when to use -> evaluation order) plus three working JSON policies validated in the IAM policy simulator

Week 2

Design Secure Architectures & Design Resilient Architectures

  • Design a multi-account landing zone with separate OU tiers for security, logging, and workloads, and place SCP guardrails at each tier

  • Build a VPC with public and private subnets, a NAT gateway path, security groups chained by reference, and NACLs, then explain why each control sits where it does

  • Select encryption mechanisms for data at rest across S3, EBS, RDS, and DynamoDB, and justify SSE-KMS versus SSE-S3 for one regulated workload

  • Choose between ACM-managed certificates, a self-signed CA, and Secrets Manager rotation for a three-tier app, stating the tradeoff for each

  • Diagram a defense-in-depth stack for a public web app that includes WAF, Shield, private subnets, and a bastion-less access path

Your checkpoint

A VPC + security architecture diagram with a written threat table listing five attack paths and the specific control that blocks each

Week 3

Design Resilient Architectures

  • Compare the four disaster recovery strategies (backup and restore, pilot light, warm standby, multi-site active-active) by computing RTO and RPO for each on the same workload

  • Configure cross-Region S3 replication and an RDS read replica failover, then record measured failover behaviour

  • Design a multi-AZ architecture for a stateless web tier and a stateful database tier, and state exactly which component fails over and how

  • Select between Route 53 failover, weighted, and latency routing for three named scenarios and defend each choice

  • Build a backup plan in AWS Backup with retention and cross-account vaulting, then verify a restore end to end

Your checkpoint

A DR runbook for one workload containing the chosen strategy, target RTO/RPO numbers, and a completed restore test with timestamps

Week 4

Design Resilient Architectures & Design High-Performing Architectures

  • Decouple a monolithic flow using SQS, SNS, and EventBridge, and pick the right one for fan-out, buffering, and event routing respectively

  • Size an Auto Scaling group with target tracking and step scaling policies, and explain how a health check type changes replacement behaviour

  • Configure an Application Load Balancer across two AZs with health checks and sticky sessions, then justify ALB versus NLB versus GWLB for three workloads

  • Choose a caching layer (ElastiCache, CloudFront, DAX) for read-heavy, global, and low-latency-database cases

  • Identify single points of failure in a supplied three-tier design and remove each one

Your checkpoint

A 40-question timed self-quiz across Domains 1 and 2 scored at 75% or higher, with every missed question rewritten as a corrected one-line rule

Week 5

Design High-Performing Architectures

  • Match EBS volume types (gp3, io2, st1, sc1) to four IOPS and throughput profiles and state the cost consequence of each mismatch

  • Select the right compute family and purchase model for a batch job, a steady-state API, and a spiky web front end

  • Design a database layer choosing between Aurora, RDS, DynamoDB, and Redshift for four access patterns, justifying partition key and index choices

  • Plan a global content delivery path using CloudFront, Global Accelerator, and Route 53 latency routing, and explain when each adds value

  • Tune a read-heavy workload using read replicas, ElastiCache, and query-level improvements, ranking them by effort versus gain

Your checkpoint

A performance design sheet mapping four workloads to compute, storage, database, and network choices with the metric that would prove each choice right

Week 6

Design High-Performing Architectures & Design Cost-Optimized Architectures

  • Instrument a workload with CloudWatch metrics, logs, and alarms, then define one composite alarm that reflects a real user-facing symptom

  • Right-size an instance fleet using Compute Optimizer or CloudWatch data and quantify the monthly saving

  • Compare Savings Plans, Reserved Instances, Spot, and On-Demand for three usage patterns and compute the blended cost

  • Build a cost allocation scheme with tags, cost categories, and budgets, then set an anomaly alert

  • Choose between serverless and always-on for a low-traffic service by calculating break-even request volume

Your checkpoint

A cost model spreadsheet for one architecture showing on-demand baseline versus optimized cost with the optimization applied line by line

Week 7

Design Cost-Optimized Architectures

  • Apply the S3 lifecycle toolkit (transition, expiration, Intelligent-Tiering) to a dataset and state the storage-class crossover points

  • Evaluate data transfer costs across AZ, Region, and internet egress, and redesign one architecture to eliminate a surprise egress bill

  • Select between NAT gateway, VPC endpoint, and PrivateLink for private access to S3 and a third-party SaaS, with the cost delta for each

  • Compare serverless pricing shapes for Lambda, Fargate, and Step Functions on a given invocation profile

  • Decide when to delete, consolidate, or reserve, and produce a decommission checklist for idle resources

Your checkpoint

A written cost-optimization review of one workload listing at least five concrete changes with an estimated monthly dollar saving for each

Week 8

Design Secure Architectures & Design Resilient Architectures & Design High-Performing Architectures & Design Cost-Optimized Architectures

  • Re-derive the four domain weighting tradeoffs and allocate your remaining review time to the weakest measured domain

  • Rewrite every missed question from earlier quizzes as a rule you can state in one sentence

  • Answer scenario questions under time pressure, focusing on requirement keywords such as 'most cost-effective', 'highest availability', and 'least operational overhead'

  • Resolve the ten concepts you flagged as uncertain by building the smallest possible working example or reading the primary documentation for each

Your checkpoint

A full-length timed mock exam scored at 80% or higher, plus a one-page personal cheatsheet of the twenty rules you got wrong most often

Milestones

Week 4: Score 75% or higher on the 40-question Domain 1 and 2 quiz, and produce a DR runbook with a restore that actually completed

Week 6: Score 75% or higher on a mixed 65-question set weighted 25/25/25/25, and produce a cost model with at least three quantified savings

Week 8: Score 80% or higher on a full-length timed mock exam with no domain below 70%

AWS Solutions Architect exam day

Your checklist

Government-issued photo ID

Bring the original, physical ID you registered with (passport or driver's license are typical). The name on the ID must match your registration exactly - if your registration name differs from your ID, fix it before exam day. Confirm the accepted ID types with the provider, since requirements vary by country and delivery method.

Second form of ID

Some test centers or jurisdictions ask for a second ID. Check your appointment confirmation from the provider rather than assuming one is enough.

Appointment confirmation and candidate number

Have your confirmation email, appointment time, and any candidate/registration ID accessible (printed or on paper, not only on a phone you may have to put away).

Test center logistics

Know the address, which entrance is open, parking or transit options, and typical security/check-in time. Do a dry run the day before if the location is unfamiliar.

Online-proctoring system check

If testing remotely, run the provider's system test on the same computer, webcam, microphone, and network you will use, at least 24 hours ahead. Confirm browser version and any required software installs.

A quiet, cleared testing space (remote)

Desk cleared of everything except permitted items, door closed, no other people or pets entering, external monitors disconnected. Proctors commonly require a room scan - confirm the exact rules in your provider's candidate handbook.

Know what is provided vs. prohibited

Expect no personal notes, books, phones, or smart devices in the testing room. Scratch materials, if any, are usually supplied by the test center - do not bring your own. Confirm permitted items with the provider.

Time budget and arrival plan

Plan to arrive well before your slot (many centers recommend ~30 minutes early for check-in) and allow buffer for traffic and ID verification. Late arrival can forfeit your seat and fee - confirm the cutoff with the provider.

Review account and current exam guide

Re-read the official exam guide for your exam code: domain weightings, question count, duration, and passing score. Version and pricing change over time - verify in your AWS Certification account.

Practice-exam score review

Bring a one-page summary of your weakest domains from practice tests (e.g., networking, cost optimization, migration) so your final review is targeted rather than a full re-read.

Logistics for the rest of the day

Clear your calendar after the exam, arrange childcare/transport, and plan a meal - testing sessions are long and you do not want to be rushed or hungry.

Plan the day

72 hours before

Confirm appointment date, time, delivery method (test center vs. online proctored), and that the name on your registration matches your ID. Reschedule now if anything is wrong - last-minute changes may incur fees or be disallowed.

72 hours before

Take one full-length timed practice exam under exam conditions. Score it, list your three weakest domains, and build a short review plan for the remaining days.

72 hours before

Shift your sleep schedule so you are alert at your actual exam start time, especially for morning slots.

24 hours before

Light review only: your one-page summary of weak areas, key service comparisons, and design trade-offs. Do not learn new services or start a fresh deep-dive topic.

24 hours before

Remote: run the system test again, update the OS/browser if required, charge devices, and confirm your internet connection. Test center: confirm address, parking, and travel time; check for local closures or transit disruptions.

24 hours before

Pack your ID, confirmation details, and directions. Set two alarms. Aim for a normal night's sleep - cramming past midnight costs more points than it earns.

Morning of

Eat a proper meal, hydrate, and avoid heavy caffeine if you are already anxious. Skim your one-page summary, then stop studying.

Morning of

Leave with buffer. For remote testing, log in early, complete check-in, close extra applications, and silence notifications and phones.

During the test

Read the full question stem and the actual ask ('most cost-effective', 'most operationally efficient', 'least operational overhead'). Take a first pass answering what you know, flag long scenario questions, and return to them.

During the test

Work at roughly 1.5-2 minutes per question; if you are stuck past that, eliminate two options, pick the best remaining, flag it, and move on.

During the test

Reserve the last 10-15 minutes to revisit flagged questions and confirm you have answered every question - unanswered questions score as wrong.

After the test

Note your provisional result if shown, then record which domains felt weak while it is fresh. Official score reports and any credential issuance follow per the provider's stated timeline.

Rules to check

  • Bring valid, unexpired government-issued photo ID with a signature; the name must match your registration exactly. Requirements differ by country - confirm with the provider before exam day.

  • Do not bring notes, books, phones, smartwatches, or other devices into the testing room. Lockers are typically provided at test centers; possession of prohibited items can void your result.

  • Follow all proctor instructions, including room scans, desk clear requirements, and remaining in webcam view if testing remotely. Do not talk to anyone or read aloud during the session.

  • Scheduled breaks are generally not offered on associate-level exams and the timer does not pause. If you leave the room or the view of the camera, the session may be terminated - confirm break policy for your specific exam with the provider.

  • Scoring is reported on a scaled basis with a passing threshold set per exam; the score report shows pass/fail and domain-level performance. Confirm the current passing score and score validity period in the official exam guide, as these can change.

  • Use the flag-and-review feature rather than leaving questions blank; unanswered questions are scored as incorrect. You may change answers before submitting.

  • Do not discuss, reproduce, or share exam content - you accept a nondisclosure agreement at the start of the session, and violations can result in score cancellation and loss of certification eligibility.

  • If a technical problem occurs mid-exam, stop and notify the proctor or support immediately before continuing; most providers have incident procedures, and unreported issues are hard to remedy afterward.

AWS Solutions Architect FAQ

How many AWS Solutions Architect practice questions are available?

The practice set has 200 AWS Solutions Architect questions. Each one includes the choices, the correct answer, and an explanation, and the set is free to read with no signup.

How many AWS Solutions Architect flashcards are available?

The flashcard set has 32 AWS Solutions Architect cards. Each card shows the term and its definition.

Which topics does the AWS Solutions Architect study material cover?

The AWS Solutions Architect study material has 4 sections, including Design Secure Architectures, Design Resilient Architectures, and Design High-Performing Architectures.

What format facts does the catalog list for the AWS Solutions Architect exam?

The catalog lists these item types: multiple choice and multiple response for the AWS Solutions Architect exam.