How to use these AWS SAA practice questions

The AWS Certified Solutions Architect (Associate) exam is a proctored, timed test with multiple-choice and multiple-response questions based on short architecture scenarios and requirements. You need to identify the constraint that matters most, choose services that meet it, and distinguish the best fit from options that are plausible but miss a requirement.

Work through each question by selecting an option, then read the explanation immediately afterwards. For every wrong answer, name the requirement it fails: availability, access control, performance, recovery, or cost. That habit is more useful than memorising service definitions without context.

1. Multi-AZ database availability

This question tests whether you distinguish database failover from web-tier scaling.

Question 11 mark

A customer-facing application runs on EC2 and uses an Amazon RDS for MySQL database. It must remain available if an Availability Zone fails. Traffic varies, and the web tier should scale without manual intervention. Which design best meets both requirements?

An Application Load Balancer and multi-AZ Auto Scaling group spread the web tier across Availability Zones. RDS Multi-AZ provides a standby for database failover. A read replica, the strongest distractor, is primarily for read scaling; it is not a substitute for Multi-AZ failover.

A high-availability question separates web-tier scaling from database failover.

When a scenario asks for both resilience and more read capacity, treat those as separate needs. Multi-AZ addresses availability; read replicas can help scale database reads. One feature should not be assumed to provide both.

2. Buffering unpredictable uploads

This question tests whether you can decouple a bursty event source from processing capacity.

Question 21 mark

Users upload images to an S3 bucket in unpredictable bursts. Each image must be processed by a Lambda function, and a sudden spike must not overwhelm processing or lose work. Which design is best?

SQS buffers bursts so Lambda can process messages at a controlled rate, and a dead-letter queue can retain messages that repeatedly fail. Direct S3-to-Lambda invocation is the strongest distractor, but it does not provide the same durable queue buffer for handling a processing backlog.

A burst-processing question tests buffering, retries and failure handling.

The key clue is that arrivals can outpace processing. A queue creates a durable boundary between upload and work; it does not remove the need to make processing safe to retry.

3. Private access to S3

This question tests route-table endpoints and whether an internet gateway or NAT device is actually needed.

Question 31 mark

EC2 instances in private subnets need to read objects from S3. They should have no internet access, and the design should avoid NAT gateway data-processing charges for this traffic. What should you configure?

An S3 gateway endpoint provides private connectivity to S3 through route tables and does not require a NAT gateway for this path. A NAT gateway is a plausible general way for private instances to reach public AWS service endpoints, but it adds a cost and is unnecessary for this S3 access pattern.

A VPC question tests the specific endpoint type used for S3 traffic.

Read the destination service carefully: S3 and DynamoDB gateway endpoints differ from interface endpoints used for many other services. A private subnet alone does not make a service request private; the route and endpoint design matter.

4. Predictable private hybrid connectivity

This question tests how the stated connection requirements guide your choice of hybrid networking service.

Question 41 mark

A company needs a private connection from its data centre to AWS. It transfers large volumes of data and requires more consistent network performance than an internet-based connection can provide. Which approach best fits?

Direct Connect provides a dedicated network connection between the data centre and AWS, fitting the requirement for predictable, high-volume connectivity. A Site-to-Site VPN is the strongest distractor for hybrid connectivity, but it runs over the internet and does not meet the stated preference for a dedicated path. A VPN can still serve as a separately designed backup.

A hybrid-network question distinguishes a dedicated connection from internet VPN connectivity.

Direct Connect should not be treated as encryption by itself. If a scenario requires encryption in transit, account for that requirement separately in the design rather than assuming the connection type provides it.

5. Secure global delivery of static content

This question tests a common combination for low-latency delivery without making the origin bucket public.

Question 51 mark

A site serves static product images from S3 to users around the world. The images should be delivered with low latency, and the S3 bucket must remain private. Which design is best?

CloudFront caches content closer to viewers, while Origin Access Control lets the distribution access a private S3 origin. Making the bucket public is the strongest distractor because it can serve content, but it fails the explicit private-bucket requirement.

A content-delivery question tests CloudFront caching and private S3 origin access.

The security condition changes the answer: do not solve a delivery problem by exposing the origin. Match the cache and origin-access design to the stated access boundary.

6. Database availability and read scaling

This question tests a two-part database requirement rather than asking you to choose only one RDS feature.

Question 61 mark

An RDS for PostgreSQL application needs automatic database failover across Availability Zones and has a read-heavy reporting workload that is slowing normal queries. Which design best addresses both needs?

Multi-AZ supports availability through standby-based failover, while a read replica can handle read-only reporting traffic. A Multi-AZ-only deployment is the strongest distractor because it addresses failover, but a standby is not the read-scaling solution requested.

A database question tests using separate RDS features for failover and read capacity.

Do not infer that a standby is automatically an application read target. Identify which component handles the normal read workload and which supports recovery.

7. Cross-Region database recovery

This question tests whether the recovery requirement calls for a cross-Region database design rather than backups alone.

Question 71 mark

A business-critical application uses Aurora PostgreSQL and must recover from a Regional outage with low recovery time and low data loss. It can run a secondary environment in another AWS Region. Which database approach best fits?

Aurora Global Database provides a cross-Region database architecture suited to low-latency replication and Regional recovery, with promotion and application recovery steps still required. A snapshot is the strongest distractor because it supports recovery, but a daily same-Region snapshot does not meet the cross-Region, low-recovery-time requirement.

A disaster-recovery question tests the difference between cross-Region replication and backups.

A cross-Region data replica is not a complete recovery plan. Include how the secondary is promoted and how the application, networking and clients reach the recovered environment.

8. Least-privilege application permissions

This question tests the identity used by an application running on ECS and the scope of its permissions.

Question 81 mark

A containerised service on Amazon ECS must read and write items in one DynamoDB table. Credentials must be temporary, and the service must not have access to other tables. What should you do?

An ECS task role supplies temporary credentials to the running task. Restrict its actions and resource to the required table. A broad task policy is the strongest distractor because it uses the right kind of identity, but it violates least privilege by granting access beyond the task's need.

An IAM question tests temporary task credentials and resource-level least privilege.

Check both parts of an IAM policy: the actions and the resources. A narrow action list attached to a wildcard resource can still grant more access than the scenario requires.

9. Shared Linux file storage

This question tests the storage type suited to multiple Linux instances mounting the same file system.

Question 91 mark

Linux application servers in several Availability Zones need simultaneous access to the same shared file system using standard file-system semantics. Which AWS service is the best fit?

Amazon EFS provides shared file storage that multiple Linux instances can mount, with mount targets in Availability Zones where clients need access. EBS is the strongest distractor because it is block storage for EC2, but a single attached volume does not provide the requested multi-instance shared file system.

A storage question tests shared file-system access across Availability Zones.

Translate the access pattern into a storage type before choosing a service: shared files, block devices and objects have different access models. Here the decisive clues are Linux file-system semantics and concurrent access.

10. Interruptible batch processing

This question tests whether workload tolerance allows the use of lower-cost, interruptible compute.

Question 101 mark

A team runs containerised batch jobs that can pause and retry if compute capacity is interrupted. Job volume varies, and the team wants managed scheduling with lower compute costs where possible. Which design best fits?

AWS Batch manages job scheduling and can use a Spot-based managed compute environment for interruptible, retryable work. A continuously running On-Demand instance is the strongest distractor because it can run the jobs, but it misses the variable demand and cost goals. Workloads must be designed for interruption and retries.

A batch-compute question tests when managed scheduling and Spot capacity fit the workload.

Review what you missed

After the set, group errors by the decision they expose. If you repeatedly confuse failover with read scaling, revisit RDS deployment patterns. If you miss private connectivity questions, trace the route from subnet to service and check whether the endpoint type matches the destination.

A board in MySummaries can group your AWS notes by architecture decision and show which areas need another pass:

Where marks go missing
58%Networking and endpoints8×
67%Database availability and DR7×
84%IAM and security6×
A topic review ranks study sections by practice performance and frequency.

For every missed question, write a one-line correction that names the requirement and the service pattern that meets it. Keep the correction specific: for example, distinguish an S3 gateway endpoint from routing S3 traffic through NAT.

A remediation card from a repeated mistake could look like this:

Remediation tray

You missed this distinction twice: which RDS feature supports standby failover across Availability Zones, and which feature is used to scale read traffic?

Add cardDismiss
A remediation prompt turns a repeated database-design error into a focused review.

How MySummaries helps

Build a revision board from your AWS notes, then use it to generate flashcards and written practice questions. Review the service patterns behind missed answers and return to the relevant material through MySummaries.