Question 470
A DynamoDB table with large overall read capacity receives most queries for one frequently viewed product ID. Throttling persists while many other partition keys are cold. Which tuning approach directly targets this specific bottleneck?
- Double provisioned write units although reads against one key are the only throttled operation.
- Create a second index with the same hot partition key and assume all load redistributes automatically.
- Increase table-wide provisioned read capacity alone without changing requests to the hot partition key or using a cache, even though other partitions are idle.
- Reduce the hot-key access rate with caching or restructure the access pattern to distribute reads where semantics allow.
Explanation
Correct Answer
D:Hot partition keys can hit local throughput limits even with spare table-wide capacity; distributing or caching requests addresses concentration.
Incorrect Answers
A:Write throughput adjustments do not correct a concentrated read hot key.
B:An index that mirrors the concentrated key can reproduce the same hotspot.
C:Extra table-wide capacity can leave the same hot partition key with localized throughput pressure; changing access concentration is the direct fix.
Question 471
A global image delivery service has already removed irrelevant analytics query parameters from its CloudFront cache policy. The remaining cache key must preserve meaningful width= and format= query values, but clients send the same key-value pairs in different parameter orders, fragmenting the cache into duplicate variants. The origin always returns identical bytes for identical pairs regardless of ordering. What edge-level change best addresses this?
- Remove width and format from the cache key entirely, so images of different dimensions and output formats are served under one cache entry.
- Reduce the distribution TTL to zero while preserving arbitrary query parameter ordering, because disabling caching increases cache reuse.
- Run a CloudFront Function on viewer requests to canonically order the retained query parameters before cache-key evaluation, while keeping width and format in the cache policy.
- Enable CloudFront Origin Shield but preserve different query string orders in viewer cache keys, assuming the shield merges those edge cache variants.
Explanation
Correct Answer
C:CloudFront Functions can normalize the query string in the viewer request, allowing equivalent parameter sets to map to the same cache key without dropping content-varying parameters.
Incorrect Answers
A:Those values change origin output; ignoring them can serve the wrong image variant and is not safe cache-key consolidation.
B:A zero TTL removes useful caching and does not normalize equivalent query strings into the same cache key.
D:Origin Shield can consolidate origin fetches from edges but does not canonicalize distinct viewer cache keys that differ only by query parameter order.
Question 472
A large Kinesis Data Stream is provisioned with 48 shards. Incoming traffic looks normal in the default stream-level CloudWatch metrics, but a small number of producers report intermittent ProvisionedThroughputExceededException. Operations must identify the exact overloaded shard IDs by shard-specific IncomingBytes, IncomingRecords and WriteProvisionedThroughputExceeded rather than guessing which partition key is hot. Which monitoring change exposes that evidence?
- Use only the existing aggregate stream metrics, which always include per-shard data even when enhanced monitoring is disabled.
- Use Kinesis enhanced fan-out consumers, because a dedicated read pipe for each consumer always reveals producer-side per-shard writes automatically.
- Enable Kinesis Data Streams enhanced shard-level CloudWatch metrics for the necessary monitored dimensions and inspect the overloaded shard IDs.
- Query CloudTrail management events for all stream operations and treat them as a one-minute per-shard IncomingBytes time series.
Explanation
Correct Answer
C:Enhanced Kinesis monitoring enables supported shard-level CloudWatch metrics, including shard-specific traffic and throttling indicators.
Incorrect Answers
A:Default stream-level metrics aggregate activity and do not provide all the requested per-shard detail without enhanced monitoring.
B:Enhanced fan-out increases supported consumer read access; it does not itself enable the requested shard-specific incoming write metrics.
D:CloudTrail provides API auditing rather than the specified shard-level CloudWatch throughput counters.
Question 473
An ecommerce OpenSearch cluster reports rising heap pressure and slow aggregations during a new analytical query. Monitoring indicates shard count is excessive for the current data size. What should the performance engineer evaluate?
- Review shard sizing, index lifecycle and query design, then consolidate excessive small shards where safe.
- Create more tiny shards on every index without checking existing heap or coordination limits.
- Increase the data-node heap allocation and instance size but preserve the existing high count of tiny shards and the same inefficient aggregation.
- Increase the number of replicas and copy every small shard to more data nodes without revisiting primary shard layout.
Explanation
Correct Answer
A:Too many small shards can add coordination and memory overhead, so right-sizing and query tuning can improve cluster resource efficiency.
Incorrect Answers
B:Shard proliferation can worsen the overhead already identified as a major contributor.
C:More heap may postpone failure but does not remove the excessive shard management and coordination work identified as the bottleneck.
D:Replicas can add search capacity but duplicate shard-management work and do not resolve a fundamental over-sharding design.
Question 474
An EC2 analytics server attaches four gp3 volumes with sufficient aggregate provisioned IOPS and throughput for measured I/O demand. Individual volumes do not reach their configured ceilings, but aggregate EBS throughput plateaus at the instance type’s documented EBS bandwidth limit. What is the most direct performance correction?
- Increase provisioned throughput on every gp3 volume while keeping the same EC2 instance whose EBS bandwidth cap is already saturated.
- Select an EBS-optimized instance type or size with sufficient aggregate EBS bandwidth, and validate combined volume and instance limits under load.
- Move only one gp3 volume to a higher provisioned-throughput setting but keep aggregate traffic on the same instance EBS connection at its measured cap.
- Add an extra gp3 volume to the existing instance without changing the EC2 EBS bandwidth limit and assume aggregate throughput grows beyond the host cap.
Explanation
Correct Answer
B:The limiting component is the EC2 instance’s EBS-optimized channel; choosing adequate host EBS bandwidth addresses the measured cap while retaining appropriately sized volumes.
Incorrect Answers
A:Volume-level throughput provisioning cannot overcome the aggregate maximum throughput enforced by the attached EC2 instance class.
C:Per-volume provisioning is irrelevant once aggregate host EBS bandwidth, rather than volume bandwidth, is saturated.
D:More attached volumes cannot increase an EC2 instance’s EBS-optimized maximum bandwidth.
Question 475
A financial trading cluster runs latency-sensitive EC2 workloads over high-bandwidth network links. Profiling shows significant packet-processing overhead inside the virtual network stack, not an application CPU-bound loop. What infrastructure feature should the architect validate?
- Disable TLS verification on client sockets and assume cryptographic bypass is the only safe network tuning.
- Enable an instance type and AMI driver with enhanced networking (ENA), and benchmark the application packet path.
- Increase attached gp3 IOPS and throughput because block-storage provisioned performance determines the instance’s packet-per-second capacity.
- Increase the MTU without verifying path support, assuming jumbo frames alone remove every packet processing bottleneck on an instance lacking enhanced networking.
Explanation
Correct Answer
B:ENA enables supported network offload and higher packet rates, making it a direct intervention for measured network-stack overhead.
Incorrect Answers
A:Disabling TLS identity verification creates security risk and does not address measured EC2 networking capacity.
C:EBS block storage throughput does not replace network adapter offload and driver features for EC2 packet processing.
D:MTU changes may help some workloads with end-to-end support but are not a substitute for the compatible ENA driver and instance networking capability.
Question 476
An EC2 scientific cluster is already using a cluster placement group in one Availability Zone, but MPI communication over ordinary TCP networking remains its measured bottleneck. The workload uses a supported EC2 instance type and MPI runtime compatible with libfabric. Which network device can reduce inter-node messaging latency further?
- Add an Elastic Fabric Adapter (EFA) on supported instances and configure the MPI/libfabric stack, then benchmark communication latency.
- Create more placement groups but continue using the exact same TCP-only network devices and MPI transport.
- Enable ENA alone, leaving the latency-sensitive MPI runtime on standard TCP and never enabling an EFA-capable device or libfabric transport.
- Add more placement group membership changes but leave the inter-node data path on the same existing TCP socket transport.
Explanation
Correct Answer
A:EFA exposes a specialized low-latency/high-throughput transport for tightly coupled MPI and HPC communication on compatible EC2 configurations.
Incorrect Answers
B:Additional placement groups do not by themselves supply the EFA network interface or MPI library integration absent from the current topology.
C:ENA improves ordinary network performance but does not supply the specialized EFA OS-bypass MPI communication requested.
D:Placement groups can reduce network distance, but the scenario already uses one and specifically calls for the further specialized EFA transport.
Question 477
A Kinesis Data Streams analytics application uses a Kinesis Client Library (KCL) worker fleet. After a worker restarts, it reprocesses hours of already analyzed stream records because its checkpoint was never advanced after successful downstream writes. Read throughput remains sufficient. Which consumer-side change fixes the recovery position?
- Register additional enhanced fan-out consumers to provide independent bandwidth but leave the existing application checkpoint stale.
- Persist KCL checkpoints after confirmed successful processing and ensure downstream work remains idempotent for any replayed records.
- Increase producer shard count to guarantee a restarted worker will start at the newest record without application checkpoint state.
- Increase Kinesis stream retention while leaving KCL record checkpoints uncommitted and assume retention controls the restart position.
Explanation
Correct Answer
B:KCL checkpoints record the application’s processed sequence position so worker restarts resume from a recent point; idempotency protects unavoidable at-least-once replay.
Incorrect Answers
A:Enhanced fan-out assigns supported read throughput per registered consumer but cannot advance an application’s checkpoint after success.
C:Shard resharding changes capacity and ownership but does not correctly persist completed processing offsets.
D:Longer stream retention protects records from expiry but does not set the last successfully processed sequence position tracked by the consumer.
Question 478
A DynamoDB-backed API executes frequent Scan operations to find a single customer by verified email address. Metrics show scanned item counts far exceed returned items, driving latency and RCUs. Which data-access redesign directly addresses the issue?
- Use FilterExpression with Scan and assume DynamoDB reads only matching indexed items before consuming capacity.
- Increase table-wide read capacity and SDK retries while preserving a Scan on each individual email lookup.
- Create a GSI partitioned only by creation-month and continue a Scan plus email filter across monthly partitions for every lookup.
- Build a suitable GSI keyed by normalized email where uniqueness and query semantics are handled explicitly, then use Query rather than Scan.
Explanation
Correct Answer
D:A designed key-based access path limits inspected items and avoids repeated full-table scans for a predictable lookup pattern.
Incorrect Answers
A:Scan filters are applied after reading items and do not turn an unindexed full scan into a key lookup.
B:Higher capacity can pay for the inefficient scans and retries but does not transform an email lookup into a targeted keyed Query.
C:A secondary index on a field unrelated to the email access pattern still fails to provide a selective email-key query.
Question 479
A business intelligence service serves repeated expensive Redshift SELECT queries to hundreds of dashboard viewers. Results change only after nightly ETL, and queries are otherwise identical. Which feature should the architect assess before adding more cluster nodes?
- Disable result cache globally while continuing to submit identical dashboards to the database.
- Configure Redshift concurrency scaling to launch additional compute for repeated identical dashboard SQL rather than checking whether a valid cached result already exists.
- Use Amazon Redshift result caching for eligible repeated queries and confirm cache reuse and invalidation behavior.
- Create a new table with a differently named copy of the report output before every dashboard refresh, even though the underlying Redshift data has not changed.
Explanation
Correct Answer
C:Eligible identical result sets can be served from Redshift query results, avoiding repeated execution until invalidated by data changes.
Incorrect Answers
A:Disabling caching removes the feature that could eliminate redundant execution for unchanged dashboard data.
B:Concurrency scaling can increase throughput for qualifying queued work, but it cannot avoid repeated execution of the same eligible query the way result caching does.
D:Repeated table copies introduce additional work and often invalidate caching rather than reusing the existing eligible query result.
Question 480
An Amazon OpenSearch Service log cluster writes event indices continuously. The cluster has already corrected its shard count and heap pressure, but large active indices keep growing beyond their planned 40-GiB per-index target, and old read-only indices should transition to a cooler tier and then expire automatically. Which control provides repeatable index lifecycle actions?
- Increase the number of primary shards for every existing index without setting a rollover or retention policy.
- Use an Index State Management policy with a write alias and supported rollover, tier-transition and deletion actions based on index age or size.
- Enable OpenSearch snapshot creation exclusively and assume backups automatically update the write alias and delete every old index.
- Move old indexes into UltraWarm manually once, without configuring any rollover conditions or future deletion policy for the growing write index.
Explanation
Correct Answer
B:Index State Management supports lifecycle policies such as rollover, warm/cold migration and deletion when age or size conditions are met.
Incorrect Answers
A:Adding shards alone does not schedule rollover, move read-only indexes across tiers or apply retention-based index deletion.
C:Snapshots protect data but do not substitute for index write alias rollover and active index lifecycle policies.
D:A one-time tier move may lower costs for eligible old data, but it does not automate requested future rollover, tier changes and retention-based deletion.
AWS Certified Solutions Architect - Professional SAP-C02 Practice Tests