AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 73
Question 720
A slow RDS report uses a full scan and a nested loop on millions of rows because a filter column has no useful index. The team proposes adding five read replicas as the first optimization. What is better?
- Add replicas and force the query to run identically on each without changing the expensive full scan.
- Increase CloudWatch log retention and assume RDS uses it as a query index.
- Replace the database's security group with a wider one to make joins avoid table scans.
- Inspect the query plan and create or tune appropriate indexes and SQL access patterns before purchasing replica capacity.
Explanation
Correct Answer
D:An inefficient query plan can remain expensive on every replica; improving data access often reduces work more directly.
Incorrect Answers
A:More hosts can spread calls, but individual queries will still perform wasteful reads.
B:Log retention does not improve relational query access paths.
C:Network authorization is unrelated to a missing relational index.
Question 721
An Aurora PostgreSQL cluster has a writer, two large read replicas reserved for checkout traffic, and two lower-cost read replicas reserved for periodic business intelligence queries. BI sessions must avoid checkout replicas without hardcoding individual instance addresses. Which endpoint design fits?
- Create an Aurora custom endpoint whose selected group contains only the two reporting replicas, and direct BI read-only connections to it.
- Use the default Aurora reader endpoint and assume it excludes the two checkout replicas because the BI SQL query text differs.
- Use the writer endpoint and assume the writer selects a smaller replica based on the SQL user’s IAM policy.
- Use a snapshot-export API as the interactive SQL endpoint for BI query sessions.
Explanation
Correct Answer
A:A custom endpoint groups specified Aurora instances and balances connection selection within that group, isolating reporting clients.
Incorrect Answers
B:The default reader endpoint can distribute connections to eligible readers generally; it does not automatically filter by client application category.
C:The writer endpoint leads to the writer; it does not route each client SELECT to a chosen specialized replica.
D:Snapshot export materializes data for offline use and is not an Aurora SQL connection endpoint.
Question 722
An Aurora PostgreSQL reporting reader repeatedly spills large GROUP BY and ORDER BY operations to temporary storage because intermediate results exceed work_mem. Query plans are already reviewed, CPU capacity is adequate, and one query’s temporary-file I/O—not the number of concurrent clients—is the bottleneck. Which targeted AWS capability should be evaluated?
- Use RDS Proxy to raise maximum application connection counts, because it stores PostgreSQL temporary relations in the proxy rather than on the database.
- Test a compatible Aurora PostgreSQL Optimized Reads-enabled DB instance with local NVMe-backed temporary objects, and monitor local ephemeral storage and query plans.
- Add more normal Aurora reader replicas but keep the affected query on the same original reader, expecting replication itself to accelerate its sort spill.
- Switch the Aurora cluster to I/O-Optimized storage pricing and assume this automatically moves PostgreSQL sort spill files onto local NVMe devices.
Explanation
Correct Answer
B:On supported Aurora PostgreSQL Optimized Reads instance classes, temporary objects and spill files can use local NVMe instead of network-backed storage, improving large sorting and aggregation workloads. Validate engine/instance support, measure query behavior and watch FreeEphemeralStorage because capacity is finite.
Incorrect Answers
A:RDS Proxy manages client-to-database connections and may help connection churn. It does not serve as PostgreSQL temporary file storage and can leave per-query sort spill I/O unchanged.
C:Extra Aurora replicas can add independent read capacity if clients distribute requests, but the same query on its existing reader still experiences the measured temporary-storage bottleneck.
D:Aurora I/O-Optimized changes the database I/O billing model; it is not the same feature as Aurora PostgreSQL Optimized Reads on supported local-NVMe instance classes. The per-query spill path must be measured and configured appropriately.
Question 723
A bursty Aurora workload has low weekday demand and brief spikes. An architect recommends Aurora Serverless without checking its engine/version support or how capacity scales during spikes. What should precede adoption?
- Assume every Aurora engine and engine version accepts every Serverless capacity configuration identically.
- Switch to DynamoDB with no schema/application design analysis solely because it is also managed.
- Validate supported Aurora Serverless engine versions, capacity boundaries, scaling behavior and observed latency requirements with a workload test.
- Set an EC2 Auto Scaling rule on the application tier and claim it automatically adjusts Aurora Serverless compatibility.
Explanation
Correct Answer
C:Serverless capacity options are conditional on supported combinations and may have minimum/maximum or scaling characteristics affecting demand peaks.
Incorrect Answers
A:Support and behavior differ by product/version and should not be invented.
B:Changing database models cannot be presumed compatible with transactional SQL queries.
D:Application scaling controls do not establish Aurora engine support.
Question 724
A DynamoDB orders table uses customer ID as partition key and sortable order date as sort key. The API needs all orders for a customer within a date range. Which data access operation is appropriate?
- Run a whole-table Scan for every user despite having the relevant partition and sort keys.
- Use GetItem with a partition key only and expect it to return all dated orders automatically.
- Use a full text SQL JOIN against DynamoDB as if it exposes a built-in relational query planner.
- Use Query with the customer partition key and a supported sort-key range condition.
Explanation
Correct Answer
D:DynamoDB can efficiently retrieve items in one partition and filter by sort-key range when the key schema matches access patterns.
Incorrect Answers
A:Scanning unrelated customers wastes read capacity and grows less efficiently.
B:GetItem requires the complete primary key and retrieves one item, not a range.
C:DynamoDB's key-value/document API does not natively execute arbitrary SQL joins.
Question 725
An order-service writes to a DynamoDB base table and immediately queries a GSI created for searches by shipment status. Occasionally the latest update is absent from the index response. Which consistency explanation is correct?
- Global secondary index reads are eventually consistent; use a suitable base-table strongly consistent read when immediate visibility is required.
- Enable ConsistentRead=true on the GSI Query and assume it guarantees strongly consistent GSI results.
- Increase GSI projection columns and expect this changes asynchronous update propagation to synchronous.
- Reduce item TTL to one second to force the newest GSI entry to appear.
Explanation
Correct Answer
A:GSI propagation is asynchronous and GSI queries cannot be made strongly consistent with the standard read-consistency option.
Incorrect Answers
B:Strongly consistent reads are not supported on DynamoDB global secondary indexes.
C:Projection selects attributes available in the index, not strong-consistency semantics.
D:TTL expiration is not a write-synchronization mechanism and can delete needed records.
Question 726
A DynamoDB system has steady predictable reads and writes each hour and can forecast its regular baseline. The architect wants predictable throughput capacity and cost management rather than unconstrained request-based billing. Which mode should be evaluated?
- Choose on-demand mode solely because it always has the lowest bill for every steady workload.
- Use provisioned capacity with measured read/write units and compatible auto scaling for demand variation.
- Disable throttling metrics and leave provisioned RCUs fixed at an unmeasured tiny amount.
- Move the DynamoDB table to EC2 instance store to avoid any read/write capacity management.
Explanation
Correct Answer
B:Provisioned mode can be economical for known usage when capacity and scaling limits reflect traffic.
Incorrect Answers
A:The cheapest mode depends on utilization and the relative pricing of provisioned and request-based units.
C:Underprovisioned fixed capacity can reject requests, undermining performance.
D:Local ephemeral storage is not a managed, durably replicated DynamoDB service replacement.
Question 727
A DynamoDB table reports high total provisioned read capacity, but requests for a single celebrity profile are throttled while other partition keys are idle. How should the performance diagnosis proceed?
- Assume available table-wide RCUs guarantee that one key can consume every partition's capacity without restriction.
- Increase the number of unused attributes stored on the hot item to raise its per-key throughput ceiling.
- Inspect hot-partition access and redesign or spread the hot key workload as appropriate, not just increase aggregate table RCUs.
- Change the DynamoDB table from provisioned to on-demand billing and assume this automatically removes all single-partition hot-key limits.
Explanation
Correct Answer
C:Per-partition throughput limits and skewed partition-key usage can throttle hot keys despite apparently adequate aggregate capacity.
Incorrect Answers
A:DynamoDB distributes data and workload among partitions; key skew can still limit one hot partition.
B:Larger items can increase consumed capacity and do not inherently remove hot-key constraints.
D:On-demand mode helps accommodate varying table throughput, but a concentrated partition key still requires attention to key distribution and applicable partition throughput limits. Billing mode alone is not a hot-key repair.
Question 728
A DynamoDB client receives ProvisionedThroughputExceededException during a temporary spike. Its retry loop immediately replays every request without delay, increasing throttling. Which client change is appropriate?
- Increase provisioned write capacity immediately while retaining an unlimited zero-delay retry loop during all future bursts.
- Depend on DynamoDB adaptive capacity and leave the caller retry storm unchanged, without inspecting hot partition keys.
- Add a DAX cache in front of the table and expect it to absorb transactional writes that are throttled by provisioned throughput.
- Use bounded exponential backoff with jitter and address capacity or partition skew based on metrics.
Explanation
Correct Answer
D:Use exponential backoff with jitter and bounded retries so concurrent callers do not synchronize repeated throttled requests. Inspect consumed capacity, hot keys and configured throughput/on-demand mode for sustained capacity problems.
Incorrect Answers
A:More capacity may address sustained aggregate pressure, but an uncontrolled synchronized retry storm still causes hot-partition pressure and amplified throttling under spikes.
B:Adaptive capacity helps some access-skew workloads but does not eliminate all throughput limits or justify unbounded immediate retries.
C:DAX is chiefly a read cache and does not remove backend write capacity constraints for updates.
Question 729
A payments application must always read the latest committed balance and cannot accept cached stale values. Engineers propose DAX as the mandatory primary read path to lower latency. Which concern matters?
- Evaluate the strong-consistency requirement against DAX's cache semantics and use DynamoDB reads that satisfy required consistency.
- Assume every DAX cached GetItem always returns the same strongly consistent result as a fresh base-table read.
- Increase DAX TTL to several hours to guarantee more immediate visibility of new payments.
- Use a GSI as an alternative strongly consistent source for the just-written item.
Explanation
Correct Answer
A:DAX accelerates cacheable reads but is not the right blanket route for strict strong-consistency guarantees.
Incorrect Answers
B:Cache behavior can return previously stored data; the consistency contract differs from a direct strongly consistent request.
C:Increasing DAX cache TTL keeps cached values eligible to be reused for longer, which may worsen staleness after a payment update. The balance requirement calls for a suitable strongly consistent DynamoDB read path instead.
D:DynamoDB global secondary index reads are eventually consistent and cannot provide that guarantee.
Question 730
A DynamoDB table scales successfully, but the DAX cache cluster saturates on read bursts and has a single point of cache failure. Which component needs its own capacity and availability plan?
- Increase DynamoDB base-table RCUs and assume this automatically creates additional DAX nodes.
- Scale and configure the DAX cache cluster with appropriate nodes and AZ redundancy independently of table read capacity.
- Disable DAX failover and use one cache node because DynamoDB itself is replicated.
- Raise the API Gateway rate limit and expect DAX to gain memory and CPU automatically.
Explanation
Correct Answer
B:DAX is a distinct service tier with its own node throughput and topology, even though it fronts DynamoDB.
Incorrect Answers
A:Table capacity and DAX cluster node allocation are separate control planes.
C:The application can still lose its cache endpoint or suffer high latency when a single DAX node fails.
D:Gateway request settings do not add DAX cache compute resources.
