Question 330
A company needs subsecond autocomplete over multilingual product descriptions and faceted filtering, while its transactional source of truth remains in Aurora. SQL LIKE searches on the writer are degrading checkout throughput. Which search architecture most directly addresses this?
- Add RDS Proxy in front of Aurora and keep full-text LIKE and faceting queries against the same transactional writer.
- Add a general-purpose Aurora read replica and continue repeated wildcard LIKE scans without an appropriate full-text search index.
- Index searchable data in Amazon OpenSearch Service using an appropriate ingestion pipeline and route search queries to the index.
- Move product descriptions into a DynamoDB table with only the order ID as a key and expect native full-text relevance scoring.
Explanation
Correct Answer
C:A purpose-built search index supports text search and aggregations without forcing the transactional database writer to scan descriptions.
Incorrect Answers
A:Proxy pooling reduces connection pressure but does not supply a scalable inverted full-text index or take text search off the writer.
B:A reader may offload some reads but does not directly implement a dedicated multilingual search and faceting workload across large descriptive text.
D:A DynamoDB item lookup table does not automatically provide ranked multilingual search, autocomplete and facets.
Question 331
An application uploads multi-gigabyte media objects to S3 from remote clients. Failed network transfers frequently force an entire large object to be sent again. The client can split files into chunks and retry failed parts. Which upload method best improves resilience?
- Enable S3 Transfer Acceleration but continue each upload as one large single PUT with no independently retried parts.
- Use S3 multipart upload with appropriately sized parts, retries and completion or abort handling.
- Retry the whole large-object PUT on failure using exponential backoff without splitting the object into independently retriable parts.
- Use S3 Batch Operations to replicate successfully completed objects instead of changing how client failures are retried.
Explanation
Correct Answer
B:Multipart upload allows individual failed parts to retry and supports parallel transfer instead of restarting every byte after a disruption.
Incorrect Answers
A:Transfer Acceleration can improve path latency but does not itself add part-level retry and resumability for a failed large upload.
C:Whole-object retries continue retransmitting successful data, which is the observed problem for unreliable large transfers.
D:Batch Operations works on existing objects and does not resume in-flight, incomplete customer uploads.
Question 332
A database application uses an EBS gp3 volume. Metrics show that bandwidth in MiB/s is saturated while provisioned IOPS remains well below the limit, and the EC2 instance supports higher EBS throughput. Which gp3 parameter should be tuned first?
- Increase only provisioned IOPS even though throughput is saturated and IOPS are already below the cap.
- Increase the EBS storage size solely because gp3 IOPS and bandwidth automatically rise with allocated GiB.
- Resize the gp3 volume for additional unused GiB and assume IOPS and throughput must rise in direct proportion to provisioned storage.
- Increase provisioned gp3 throughput within volume and instance limits and validate the workload with metrics.
Explanation
Correct Answer
D:gp3 lets throughput and IOPS be tuned independently; the measured ceiling is bandwidth rather than operation count.
Incorrect Answers
A:Additional IOPS cannot fix a separate throughput ceiling when average request size drives MiB/s saturation.
B:gp3 performance is configurable independently of size for supported settings; adding unused GiB is not the targeted fix.
C:gp3 throughput is separately configured within supported limits; adding unused storage does not directly raise its provisioned MiB/s ceiling.
Question 333
A global retail site already uses CloudFront, but duplicate requests for a popular image converge on regional edge caches and repeatedly hit the S3 origin during cache misses. Which CloudFront feature can reduce origin load from multiple edges?
- Evaluate Origin Shield at an appropriate Region and cache configuration to consolidate eligible origin fetches.
- Disable CloudFront regional edge caching for popular objects and rely on every edge cache to independently refetch from S3.
- Configure a second CloudFront origin solely for 5xx failover while leaving successful cache misses from many edges independent.
- Make the cache key include each request ID and tracking parameter so identical images remain separate across edge locations.
Explanation
Correct Answer
A:Origin Shield adds another caching layer that can improve the consolidated cache-hit behavior for requests arriving from multiple edge locations.
Incorrect Answers
B:Removing shared caching makes origin fetch duplication worse rather than consolidating cache misses.
C:Origin failover addresses origin unavailability; it does not by itself consolidate routine cache misses at the primary origin.
D:A cache key with irrelevant per-request entropy lowers hits and increases rather than reduces origin fetch frequency.
Question 334
A company serves a public REST API from two healthy regional Application Load Balancers. Clients can tolerate DNS caching and do not require static anycast IPs. Operations wants Route 53 to steer DNS answers to the AWS Region with the lowest observed latency for each resolver and stop using an unhealthy target. Which record policy is the best starting point?
- Use Route 53 latency-based routing records for both regional endpoints, with suitable health evaluation and TTL planning.
- Use only one ordinary A record pointed permanently at the first Region regardless of latency and endpoint health.
- Require Global Accelerator static anycast entry points even though the requirement specifically calls for DNS-response-based regional steering.
- Use Route 53 geolocation records that select a fixed country-to-Region mapping regardless of actual measured network latency.
Explanation
Correct Answer
A:Route 53 latency routing selects eligible regional resources based on measured network latency and associated health state when configured; resolver caching and DNS TTL remain operational factors.
Incorrect Answers
B:A fixed single-region record does not express two-region latency-based placement or failover.
C:Global Accelerator is an alternative for static anycast and accelerator routing but is not necessary to satisfy this expressly DNS-driven solution requirement.
D:Geolocation routing uses location rules, not the latency selection stated for healthy regional ALBs.
Question 335
A DynamoDB application supports a global celebrity product. One partition key accounts for most traffic and returns throttling even when table-level capacity is sufficient. The application can cache frequent immutable reads. What performance strategy best addresses the specific skew?
- Cache hot product reads and redesign partition-key access as needed so a single key is not the sole throughput bottleneck.
- Increase table provisioned read capacity evenly without changing the requests that all target one hot partition key.
- Create a GSI with the same high-traffic partition key and route all the identical hot-product reads through it.
- Create a DynamoDB Streams consumer that copies each hot read request to a logging stream.
Explanation
Correct Answer
A:Per-partition hot-key concentration can cause throttling despite spare table capacity; spreading or caching targeted operations reduces localized pressure.
Incorrect Answers
B:A single physical hot partition/key can remain throttled even if account-wide or table-level capacity rises.
C:Preserving the same concentrated partition-key access on another index can reproduce skew rather than spread or cache it.
D:Streams capture item change records, not a hot-key read cache or partition throughput mitigation.
Question 336
An analytics lake issues many Athena queries over JSON records in S3. The same aggregate is recomputed every hour, most queries touch only recent date partitions, and engineers measure very high bytes scanned. Which layout improvement is most appropriate?
- Convert relevant datasets to columnar compressed Parquet and partition on useful date predicates with compact files.
- Retain small uncompressed JSON objects and update the Glue catalog frequently without changing the scanned columns or partition layout.
- Purchase Athena provisioned capacity solely to reduce the scanned-byte count of unfiltered JSON queries.
- Enable S3 Intelligent-Tiering on all JSON objects and expect the Athena engine to scan fewer bytes per query without query or format changes.
Explanation
Correct Answer
A:Columnar formats and partition pruning reduce unnecessary scanned bytes and file open overhead for appropriate Athena queries.
Incorrect Answers
B:Catalog refresh does not reduce the bytes scanned from verbose row-oriented objects when queries touch many files.
C:Provisioned capacity addresses query compute isolation and utilization; it does not turn row-oriented JSON into a pruned columnar layout.
D:Storage-tier pricing may change storage bills but does not itself reduce Athena scanned-column bytes.
Question 337
An ECS service on Fargate scales desired task count from 20 to 80 during a traffic peak, but new tasks fail to start with an error allocating network interfaces in a designated subnet. Existing tasks have normal CPU utilization, and the subnet has almost no unused private IPv4 addresses. What infrastructure change addresses the observed limit?
- Increase ECS service desired count to 160 in the same exhausted subnet without enlarging its remaining private IP capacity.
- Add enough IP-address capacity through appropriately sized subnets and permitted AZ subnet configuration for awsvpc task ENIs, then test task placement.
- Increase each Fargate task ephemeral-storage allocation while leaving the subnet IPv4 pool exhausted.
- Switch to Docker bridge network mode for Fargate and assume multiple tasks can share one ordinary host interface.
Explanation
Correct Answer
B:Fargate tasks use task network interfaces with subnet IP addresses; sufficient IP capacity and subnet selection are prerequisites to scaling.
Incorrect Answers
A:More desired tasks cannot launch once the selected Fargate awsvpc subnet has no usable IP addresses.
C:Task scratch disk allocation has no relation to IP addresses assigned to task ENIs.
D:Fargate requires supported awsvpc task networking; bridge mode is not a way around subnet ENI address exhaustion.
Question 338
A global API serves personalized HTML behind CloudFront. Its cache-hit ratio is poor because every request forwards a random analytics header that does not change the response, while authentication cookies must remain isolated. Which CloudFront configuration distinction matters?
- Use an appropriate cache policy that includes response-varying inputs while excluding irrelevant analytics headers from the cache key, with secure authorization handling.
- Include every randomly generated header in the cache key because a unique key always maximizes cache reuse.
- Strip all cookies including those controlling customer authorization and cache personalized bodies globally.
- Forward the analytics header using an origin request policy but also explicitly include it in the cache key for every personalized response.
Explanation
Correct Answer
A:Cache key cardinality determines edge reuse; omitting irrelevant headers can improve hits without merging genuinely personalized results.
Incorrect Answers
B:Unnecessary key variation fragments the cache into entries that cannot be reused across identical responses.
C:Merging authorized customer responses can expose private content and breaks application semantics.
D:Forwarding an irrelevant field to the origin need not put it into the cache key; including it still fragments otherwise reusable entries.
Question 339
A CloudFront distribution fronts two private REST origins. The team wants to forward an Authorization header to an origin for verification, but does not want it included in the cache key for an endpoint configured to disable caching. Which policy separation should be reviewed?
- Add Authorization only to the cache-key policy and simultaneously require that it never contribute to the cache key.
- Attach a CloudFront response headers policy for Authorization instead of forwarding the viewer request header to the custom origin.
- Use only CloudFront function code to modify a viewer URL while leaving all policies to discard Authorization on origin requests.
- Configure origin request policy to forward required headers while using an appropriate cache policy for the endpoint’s response semantics.
Explanation
Correct Answer
D:Origin request policies control forwarded values; cache policies define cache-key variation and caching behavior.
Incorrect Answers
A:A header included in the cache policy becomes a cache-key dimension; origin request policy is the separation required when caching is disabled.
B:Response headers policies act on the response to the viewer rather than controlling which viewer credentials reach the origin.
C:Rewriting the URI does not cause CloudFront to retain an Authorization header that its forwarding policy omits.
Question 340
A CloudFront-backed site serves large text and JSON objects. Trace data shows network transfer time dominates for clients on slower mobile links, and responses are compressible. Which CDN optimization should be considered?
- Enable Origin Shield for every response while keeping the same uncompressed text payload size for each mobile viewer.
- Increase ALB backend keep-alive and session stickiness while keeping JSON payloads unchanged.
- Enable supported CloudFront automatic compression or origin compression for eligible content types and verify cache behavior.
- Select a different CloudFront price class solely to shrink the uncompressed JSON returned to each viewer.
Explanation
Correct Answer
C:Compression reduces bytes transferred for eligible text responses, improving latency on constrained client networks.
Incorrect Answers
A:Origin Shield can reduce origin misses but does not compress the bytes transferred to a constrained viewer.
B:Connection reuse can reduce setup overhead, but the stated latency is dominated by content bytes and compressibility.
D:Edge-region cost selection does not directly enable content encoding or reduce response size.
AWS Certified Solutions Architect - Professional SAP-C02 Practice Tests