AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 51
Question 500
A product-catalog API repeatedly queries the same 200 most popular SQL rows. The RDS writer has adequate connection headroom but high CPU from identical read queries; prices may be up to 30 seconds old, but updates must eventually be reflected. Which approach reduces repeated SQL execution?
- Increase RDS Proxy client connections and assume it caches SQL result sets for repeat reads.
- Replace the catalog read endpoint with an RDS automated snapshot, which remains an interactive SQL connection to current prices.
- Add an ElastiCache Redis-compatible cache-aside layer keyed by product ID with a short TTL and update-aware invalidation.
- Turn on S3 Object Lock on an unrelated bucket so SQL SELECT requests are served from immutable objects.
Explanation
Correct Answer
C:Cache-aside reads from the cache first and populates missing entries from RDS; TTL and invalidation enforce the allowed staleness window.
Incorrect Answers
A:RDS Proxy pools connections and manages failover; it does not provide a SQL result cache for identical catalog reads.
B:Automated snapshots are backups, not a low-latency query endpoint serving live relational rows.
D:Object Lock protects S3 object retention and cannot intercept an RDS SQL API for caching.
Question 501
An S3 ingestion workflow receives 1,000 files per second at peak. A Lambda parser can process only 100 per second and direct object-created invocations overwhelm a downstream database. What buffer should be introduced?
- Route eligible S3 events through SQS and scale the parser fleet against a durable queue backlog.
- Run one synchronous S3 ListObjects scan for every file upload and assume it improves burst absorption.
- Remove the downstream database's connection cap without checking its actual compute or transaction capacity.
- Invoke all parser functions directly without concurrency limits because Lambda automatically increases the database's maximum connections.
Explanation
Correct Answer
A:The queue smooths spikes and supports bounded worker concurrency while retaining work until downstream capacity catches up.
Incorrect Answers
B:Polling does not create a durable processor backlog and adds unnecessary S3 request work.
C:An arbitrary higher cap may worsen saturation instead of safely smoothing the work arrival rate.
D:Serverless compute scaling can overwhelm a finite database unless work is buffered or throttled.
Question 502
An object-transformation Lambda retries a malformed source file repeatedly and eventually fails permanently. The business wants its filename and failure context retained for later redrive. Which failure-routing architecture should be configured?
- Disable application exception reporting so the invocation appears successful even if no transformed output exists.
- Use an appropriate DLQ or failure destination along the event-driven processing path, with controlled redrive procedures.
- Set unlimited retries and never isolate the bad file, even when its input is permanently invalid.
- Delete every malformed source object before capturing any evidence, then declare the processing result complete.
Explanation
Correct Answer
B:Durable failure routing preserves rejected work when bounded retries cannot succeed, enabling diagnosis without blocking healthy ingestions.
Incorrect Answers
A:False success discards processing state and deprives operators of the failure information needed for recovery.
C:Infinite retries consume capacity while deterministic failures remain unresolved and can delay unrelated valid work.
D:Destructive deletion obscures the cause and prevents restoring work once the converter is repaired.
Question 503
An S3 object-event consumer records document processing results in DynamoDB. Duplicate ObjectCreated notifications sometimes cause the same version to be charged for extraction twice. What should be recorded?
- Use the current Lambda invocation request ID as the only unique processing key for each retry of the same object.
- Remove object version IDs from records because S3 guarantees exactly one notification for every successful PUT.
- Track object identity and version or a stable processing identifier, and use idempotent conditional writes before charging.
- Increase S3 Transfer Acceleration to prevent retries of unrelated object notifications.
Explanation
Correct Answer
C:Event deliveries can repeat. Storing a deduplication key with an atomic claim prevents repeated business side effects for one source object version.
Incorrect Answers
A:Every retry can have a new invocation ID, so it does not reliably recognize the previously processed object.
B:Duplicate delivery is possible, and ignoring version identity can corrupt outcomes when keys are overwritten.
D:Transfer endpoint optimization does not provide exactly-once event delivery to processors.
Question 504
A document-processing application uses an SQS FIFO queue as the input to an existing serialized workflow. A new S3 general-purpose bucket must send Object Created notifications into that queue. An architect tries to configure the FIFO queue ARN directly in the S3 event-notification destination field but the configuration is rejected. What is the supported event-routing design?
- Change only the FIFO queue content-based deduplication setting; once deduplication is enabled, S3 direct notifications can target that FIFO queue ARN.
- Use an S3 bucket replication rule to deliver object-created notifications as FIFO queue messages, retaining the destination queue as a replication target.
- Configure an SNS FIFO topic as the direct S3 event-notification destination and subscribe the SQS FIFO queue to it, without using EventBridge.
- Enable S3 event delivery to Amazon EventBridge and create a matching rule targeting the SQS FIFO queue with suitable message group, deduplication configuration and resource permissions.
Explanation
Correct Answer
D:Amazon S3 supports EventBridge delivery, and EventBridge rules can target SQS FIFO queues with the required queue content-based deduplication and SqsParameters message group. Delivery to a FIFO queue does not independently prove chronological ordering of the source S3 events.
Incorrect Answers
A:Changing FIFO deduplication alters message handling in SQS; it does not change the Amazon S3 event-notification destination restriction, which excludes direct SQS FIFO queue targets.
B:S3 replication asynchronously copies eligible object data to supported buckets, not to SQS queues as an event delivery mechanism. It cannot transform a queue into an S3 replication destination.
C:Amazon S3 direct event-notification destinations support SNS Standard topics but not SNS FIFO topics. Although an SNS FIFO topic can have an SQS FIFO subscription, S3 cannot directly publish notifications to that topic as proposed.
Question 505
A manufacturing partner can upload reports only through a legacy SFTP client and cannot implement AWS S3 APIs. The cloud solution should avoid maintaining custom SFTP servers. Which AWS service handles this protocol ingress?
- Use AWS Transfer Family with a managed SFTP endpoint and appropriate S3 or EFS backend.
- Create an SQS queue name ending in .sftp and assume it accepts SSH file-transfer connections.
- Require the partner to send SFTP commands directly to an S3 HTTPS object endpoint without any translation layer.
- Use CloudFront HTTPS cache invalidation as an SFTP server for every authenticated partner session.
Explanation
Correct Answer
A:Transfer Family provides managed file transfer protocols compatible with partner clients while integrating authenticated access to AWS storage.
Incorrect Answers
B:Queue naming cannot change SQS into a managed SSH/SFTP file server with protocol support.
C:Native S3 object APIs do not speak the SFTP network protocol used by the existing client.
D:CDN cache invalidation is not an interactive SFTP file-transfer endpoint or protocol adapter.
Question 506
An SFTP upload of a 20-gigabyte partner file triggers a slow validation pipeline. The partner connection should end promptly once transfer storage is complete; validation can continue for an hour. Which architecture decouples the two operations?
- Perform all validation synchronously inside the SFTP session and keep the partner TCP connection open until it finishes.
- Store file bytes only in the SFTP server process's temporary memory after closing the connection.
- Route validation through Route 53 DNS failover only and do not create a processing work item.
- Land the file through Transfer Family into S3 or EFS, then schedule validation asynchronously from stored-file events.
Explanation
Correct Answer
D:A durable file landing zone separates network transfer lifetime from independent processing, so partner sessions do not wait for downstream computation.
Incorrect Answers
A:Hour-long processing attached to a transport session increases timeout and retry exposure unnecessarily.
B:Ephemeral memory is not a safe durable landing zone for large partner datasets or replay on processor failure.
C:DNS availability does not schedule or persist a file-validation job after a completed upload.
Question 507
Several external suppliers upload to one Transfer Family SFTP endpoint. Each supplier must access only its own storage prefix and must not see other suppliers' files. What identity/authorization approach should be used?
- Make the entire landing bucket public-read and public-write, then tell partners not to inspect others' folders.
- Map supplier identities to appropriately restricted storage access roles and logical directories or prefixes.
- Use a single unrestricted IAM role for all suppliers and rely only on client-visible folder names.
- Issue the AWS account root access key to every supplier so all can manage permissions themselves.
Explanation
Correct Answer
B:Transfer Family supports authenticated identities associated with scoped underlying storage permissions, isolating partners sharing one endpoint.
Incorrect Answers
A:A written agreement is not a technical boundary against unauthorized reads and overwrites.
C:An unscoped role can still access objects outside a displayed directory unless backend authorization enforces isolation.
D:Shared root credentials break tenant isolation, accountability and least-privilege storage access.
Question 508
A partner retries a timed-out SFTP upload, creating two arrival events for the same business document. The downstream ETL job could process the document twice. What recovery design is needed?
- Use only longer DNS TTLs to keep retries from generating duplicate downstream jobs.
- Assume every retry creates a new legitimate document and charge the partner twice for each same-content upload.
- Track transfer completion and stable document identity, then use idempotent asynchronous processing across retries.
- Disable SFTP transfer logging and treat every duplicate processed file as an unrelated upload.
Explanation
Correct Answer
C:The transport retry and the processor's job lifecycle are separate. A stable deduplication record prevents duplicate ETL side effects after retransmission.
Incorrect Answers
A:DNS caches do not impose an application deduplication key or control processing of repeated file-arrival events.
B:Transport timeouts do not necessarily mean the original file was lost; duplicate submissions can represent one business document.
D:Hiding transfer evidence makes recovery harder and does not establish exactly-once business processing.
Question 509
A fleet of stateless web servers scales across three Availability Zones. Developers place a shared user-upload directory on the EBS volume attached to the first server and expect all other servers to mount that exact volume concurrently across AZs. What design issue exists?
- EBS volumes automatically become globally shared read-write filesystems when EC2 Auto Scaling launches a second instance.
- Ordinary EBS block volumes are bound to an AZ and do not provide transparent multi-AZ shared POSIX file access.
- Choose an ALB sticky-session setting; it makes the same EBS block device simultaneously accessible across Regions.
- Enable S3 Transfer Acceleration on the EBS volume to mount it as a distributed POSIX filesystem.
Explanation
Correct Answer
B:EBS is block storage with AZ attachment restrictions. Shared multi-AZ filesystem semantics require a service designed for that role, such as EFS.
Incorrect Answers
A:Auto Scaling does not transform AZ-scoped block devices into a multi-host distributed filesystem.
C:Request affinity cannot alter block volume attachment or distributed filesystem semantics.
D:S3 transfer settings do not apply to EBS volume attachment or multi-AZ file-sharing.
Question 510
An application requires POSIX file locking and one shared media workspace mounted by EC2 instances across multiple Availability Zones. Teams want the shared storage to remain available when an instance is replaced. Which managed AWS file storage fits?
- Use Amazon EFS with appropriate mount targets and access controls across the required Availability Zones.
- Use S3 object PUT requests as a drop-in POSIX shared file system with standard lock semantics.
- Keep assets only in one EC2 instance-store directory and route all users to that instance forever.
- Use a separate ordinary EBS volume on each instance and assume all writes replicate instantly between volumes.
Explanation
Correct Answer
A:EFS provides managed network file-system semantics and can be accessed from multiple supported EC2 instances rather than binding data to one host's block device.
Incorrect Answers
B:Object storage has different consistency and filesystem semantics than a mounted NFS shared file service.
C:Instance-store data is ephemeral and not a resilient, multi-AZ shared filesystem.
D:Independent EBS volumes do not automatically synchronize a shared POSIX workspace or file locking.
