AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 5
Question 40
A manufacturer must repeatedly synchronize tens of terabytes of files from an on-premises NAS to an S3 bucket. It wants scheduled transfers, validation and managed throughput handling rather than operating custom rsync hosts and restart scripts. What is the best fit?
- Use S3 static website hosting to copy files from SMB/NFS automatically without transfer agents.
- Use CloudTrail event history as a managed file-replication engine.
- Attach one EBS Multi-Attach volume across the on-premises network and to the S3 bucket.
- Configure AWS DataSync with a supported source location, destination, agent/connection model and scheduled transfer tasks.
Explanation
Correct Answer
D:DataSync offers managed transfer tasks between supported file/object locations, with scheduling, parallelism and verification capabilities. The network path and endpoint prerequisites must be sized for the data volume.
Incorrect Answers
A:S3 website hosting serves public website content but cannot crawl an on-premises NAS and continuously perform scheduled, verified bulk copy jobs.
B:CloudTrail records AWS API activity; audit logging is not a managed payload-movement and checksum-validation mechanism for NAS files.
C:EBS is a zonal block service attached to supported EC2 instances, not a distributed mount bridging local NAS storage into an S3 bucket.
Question 41
A Java microservice has been tested successfully on arm64, and performance benchmarks show better price-per-request on a Graviton-based EC2 family than the current x86 family. Which migration plan best uses those findings?
- Migrate the compatible workload to a suitably sized Graviton instance family after validating binaries, dependencies and production performance.
- Retain x86 indefinitely because EC2 cannot run Java applications on an ARM processor.
- Replace the service with Amazon Athena because it provides an instruction-set-compatible EC2 CPU.
- Increase the CloudWatch metrics retention period to change the EC2 instruction set automatically.
Explanation
Correct Answer
A:Graviton instances use the ARM architecture and can offer favorable price-performance for compatible workloads. Verified arm64 dependencies and representative performance tests make the choice defensible; rollout should still be gradual.
Incorrect Answers
B:Java runs on supported ARM64 runtimes, subject to native libraries and deployment dependencies. An unconditional claim of incompatibility contradicts the stated successful compatibility tests.
C:Athena is a serverless SQL query service, not an EC2 processor upgrade or host replacement for a running Java API.
D:Metrics retention affects observability storage, not the CPU architecture or purchase economics of the EC2 instances.
Question 42
An EC2 Auto Scaling group behind an ALB handles unpredictable request volume. Operators want average requests per target to stay near a chosen level as traffic rises and falls, without repeatedly editing an hourly schedule. Which scaling approach is best?
- Configure one fixed scheduled scale-out per day regardless of actual request rate.
- Raise the minimum group size permanently to the observed peak and disable scaling policies.
- Configure target-tracking scaling using an appropriate ALB request-count-per-target metric and a tested target value.
- Apply an S3 bucket lifecycle rule to scale EC2 based on object storage age.
Explanation
Correct Answer
C:Target tracking automatically adjusts desired instance capacity around a target value of a supported utilization/workload metric. ALB requests per target is suitable when correlated with per-instance work.
Incorrect Answers
A:Fixed schedules can help predictable demand but do not react appropriately to unexpected, irregular traffic in this scenario.
B:Permanently running peak fleet capacity wastes resources during quiet periods and does not adapt to a higher unanticipated peak.
D:S3 object lifecycle rules change object storage classes or expiration, not EC2 Auto Scaling desired capacity based on web requests.
Question 43
A DynamoDB table already uses on-demand capacity. Its current warm throughput is much lower than an announced tenfold write surge at a scheduled game launch. The application must handle the first-minute peak without waiting for organic warm-throughput growth, and the team wants to keep on-demand billing. What should the architect prepare?
- Add a DynamoDB Accelerator cache to guarantee that every new write can exceed the table’s warm write-throughput value from the first minute.
- Request sufficient DynamoDB table and relevant GSI warm-throughput pre-warming before launch, allow the operation to complete, and confirm account/Region capacity quotas while retaining on-demand mode.
- Raise only the DynamoDB on-demand maximum-throughput ceiling while leaving its warm-throughput baseline below the launch's initial write rate.
- Switch to provisioned capacity with reactive target-tracking Auto Scaling just before the launch, then return to on-demand mode afterward.
Explanation
Correct Answer
B:DynamoDB warm throughput is the immediately available request rate for a table or index. AWS supports proactive paid pre-warming for anticipated spikes without switching capacity mode. Plan sufficient time and check quotas and any affected GSI warm values, while also distributing keys to avoid hot partitions.
Incorrect Answers
A:DAX caches eligible reads but does not increase a table’s underlying write throughput. Caching cannot guarantee the requested immediate write-ingestion capacity.
C:The on-demand maximum throughput cap limits admitted request rates but does not itself pre-warm table or GSI capacity for the first minute of a sudden spike. The already-insufficient warm throughput remains a bottleneck.
D:This changes the required billing mode and relies on reactive scaling, which can lag the first-minute demand. Proactive warm-throughput adjustment is supported while retaining on-demand mode.
Question 44
A gaming service repeatedly reads the same hot DynamoDB keys and can tolerate eventual consistency for those reads. Benchmarks show that application-side query caching is the most effective next step. Which managed option provides a DynamoDB-oriented in-memory cache?
- Add DynamoDB global tables solely to cache the same keys in the primary Region.
- Move every read to S3 Glacier Flexible Retrieval for microsecond access.
- Increase DynamoDB TTL expiration speed and expect expired items to be returned from memory.
- Evaluate DynamoDB Accelerator (DAX) for supported eventually consistent reads and design its cache sizing and availability.
Explanation
Correct Answer
D:DAX provides a managed DynamoDB-compatible in-memory cache for supported read patterns. It can reduce repeated-read latency when the application's consistency expectations allow cached, eventually consistent results.
Incorrect Answers
A:Global tables replicate DynamoDB data across Regions for regional access and availability. They do not install an in-memory DAX cache for hot keys in the same Region.
B:Glacier Flexible Retrieval has archive retrieval delays and is not a microsecond key-value read cache for DynamoDB.
C:TTL marks records for asynchronous deletion; it does not implement a read-through in-memory cache or accelerate recently accessed values.
Question 45
A company connects dozens of application VPCs and requires controlled transitive connectivity through central inspection and hybrid networking. The current full-mesh VPC peering design is becoming difficult to manage. What architecture best addresses the routing scale?
- Use a separate nontransitive VPC peering link for every possible pair and avoid shared routing policy.
- Put an internet gateway in each VPC to create automatic private transitive routes between them.
- Adopt AWS Transit Gateway attachments and route tables for controlled hub-and-spoke and supported transitive VPC/hybrid routing.
- Use IAM Identity Center permission sets as a replacement for IP routing between VPC CIDR blocks.
Explanation
Correct Answer
C:Transit Gateway serves as a regional network transit hub for attached VPCs and other supported networks. Separate route tables and attachments can implement the requested traffic segmentation and central paths.
Incorrect Answers
A:Peering is nontransitive; a full mesh requires many direct connections and separate routes, which is exactly the operational scale problem described.
B:An internet gateway enables internet routing for eligible VPC traffic but does not create private transitive routing between unrelated VPCs.
D:IAM Identity Center governs workforce identity and permissions, not packet forwarding, VPC attachment topology or route propagation.
Question 46
A startup hosts api.example.com and app.example.com on the same HTTPS Application Load Balancer. Both services use the root path `/`, but each must forward to a different ECS target group based on the requested hostname. Which listener condition avoids mixing those services?
- Configure ALB HTTPS listener rules using host-header conditions for api.example.com and app.example.com, each forwarding to its own target group.
- Configure path `/` as the sole matching condition for both rules and expect it to distinguish the requested hostnames.
- Use an EBS volume tag on ECS tasks to route incoming HTTPS traffic by the browser hostname.
- Route all requests to one target group and assume the load balancer reads the application response to choose a second group.
Explanation
Correct Answer
A:ALB listener rules can match the HTTP Host header and forward requests to different target groups even when both applications use the same URL path. The target groups can scale independently.
Incorrect Answers
B:Both applications share the same path, so a path-only rule does not distinguish their hostnames.
C:EBS storage tags are not ALB listener-routing conditions.
D:Listener forwarding occurs before the target serves the request, not in response to its content.
Question 47
A data lake receives new CSV files each night. A team needs repeatable schema-aware transformations into Parquet and catalog updates, but does not want to maintain a permanently running Spark cluster. Which managed service is a good fit?
- Use Route 53 latency routing to rewrite CSV columns before queries.
- Use AWS Glue jobs and catalog capabilities to transform and register the resulting datasets, with job scheduling and error handling.
- Use AWS WAF rules to convert streaming file formats at the application firewall.
- Use S3 Object Lock compliance mode to infer database schema and execute joins automatically.
Explanation
Correct Answer
B:AWS Glue can run managed ETL workloads and use the Glue Data Catalog to describe datasets. It supports the objective without teams operating an always-on Spark cluster, while requiring compatible scripts and capacity choices.
Incorrect Answers
A:Route 53 chooses DNS answers and cannot parse, transform or catalog rows from data lake files.
C:WAF inspects HTTP requests for filtering; it does not run bulk data transformation jobs or write Parquet files from CSV input.
D:Object Lock enforces version-level retention when enabled correctly. It has no transformation engine or schema inference feature for a data pipeline.
Question 48
Business analysts need occasional SQL reports over partitioned Parquet files already stored in S3. They prefer a serverless query model and do not need a continuously running transactional database. Which service should they consider?
- Deploy one large RDS Multi-AZ instance solely to scan files still located in S3 without importing or querying them.
- Use SQS long polling to execute SQL SELECT queries across partitioned S3 columns.
- Use Elastic Load Balancing to translate HTTP requests into Parquet scans.
- Use Amazon Athena with appropriate catalog tables, partitions and workgroup controls to query the S3 datasets directly.
Explanation
Correct Answer
D:Athena executes SQL directly against supported datasets in S3 without provisioning query clusters. Partitioning and columnar formats can reduce scanned data and improve cost and performance.
Incorrect Answers
A:An RDS database runs SQL over its own engine-managed data; provisioning it does not automatically query external Parquet files in S3 and introduces idle compute management.
B:SQS long polling is a queue receive feature, not an analytical SQL execution engine for objects stored in S3.
C:A load balancer distributes supported connection or HTTP traffic, not serverless SQL statements over columnar files.
Question 49
A company stores millions of medium-sized S3 objects. Access frequency varies unpredictably over the year, and the objects must remain available with automated placement as usage cools. The team accepts applicable monitoring fees. Which option best reduces manual storage-class decisions?
- Move all objects to S3 Glacier Deep Archive immediately even though users need unpredictable prompt access.
- Place all objects in S3 One Zone-IA although the business requires multi-AZ resilience for this dataset.
- Evaluate S3 Intelligent-Tiering with eligible objects and the applicable monitoring/automation charges.
- Create an EBS root volume for every object so each item inherits gp3 block-storage tiers.
Explanation
Correct Answer
C:S3 Intelligent-Tiering can move qualifying objects among access tiers based on changing observed access, reducing the need to forecast when each individual object becomes cold. Its request/monitoring economics must be modeled.
Incorrect Answers
A:Deep Archive is optimized for infrequent archival retention with restore delay and would conflict with unpredictable prompt retrieval requirements.
B:One Zone-IA stores data in one Availability Zone and is unsuitable when the stated dataset requires multi-AZ resilience.
D:EBS is block storage for attached compute and does not supply per-object S3 storage-class monitoring and automated tier movement.
Question 50
A multi-AZ analytics application uses an EFS file system. Most files are read intensively for one week and then rarely accessed for months. The business wants to retain a normal shared filesystem interface while lowering cold-data storage cost. What should it evaluate?
- Configure appropriate EFS lifecycle management to move eligible infrequently accessed files into a lower-cost EFS storage class, accounting for access charges.
- Migrate cold files into S3 Glacier Flexible Retrieval, but make no changes to the existing NFS clients or their file access paths.
- Create scheduled AWS Backup recovery points for EFS and shorten backup retention while leaving all live files in the EFS Standard storage class.
- Switch EFS to Elastic throughput mode and rely on the throughput setting to automatically reclassify the old files as infrequently accessed storage.
Explanation
Correct Answer
A:EFS lifecycle management can shift eligible cold files into its infrequent-access tiers while retaining the EFS shared-filesystem model. Retrieval/access charges and latency tradeoffs should be considered.
Incorrect Answers
B:S3 Glacier Flexible Retrieval is an archival object-storage class rather than transparent EFS lifecycle tiering. Its restore and object-access workflow will not keep unchanged NFS mounts serving those files.
C:Backup recovery point retention controls backup copies, not the storage class of cold live files inside the active EFS file system. Its main Standard storage footprint would persist.
D:Throughput mode determines filesystem throughput economics, whereas lifecycle policies decide when eligible file contents transition into EFS IA or Archive. Changing throughput mode alone does not move cold data.
