Amazon AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 70

Question 690

An EC2 rightsizing recommendation would cut node capacity in half. The application has quiet weekends, rare month-end spikes and a memory-heavy batch. Select TWO measurements required before applying the change.

Apply rightsizing from only the minimum overnight CPU sample and omit month-end memory, disk and network demand measurements.
Load-test or stage the proposed smaller instance under the target SLOs before replacing production capacity.
Measure representative peak CPU, memory, disk and network needs including the month-end batch.
Review only the lowest overnight CPU utilization and ignore all other dimensions.
Discard recent memory-pressure and swap alerts to remove noisy data from the rightsizing recommendation.

Explanation

Correct Answer

A candidate size recommendation is not proof of acceptable worst-case latency or out-of-memory behavior.

EC2 instance suitability depends on more than average CPU; month-end working sets may dominate RAM or I/O.

Incorrect Answers

Rightsizing based on idle CPU can violate peak memory/I/O and application SLOs; representative peak metrics and validation are required.

A low idle CPU sample cannot establish enough RAM or throughput for rare production peaks.

Excluding adverse evidence would hide the exact bottleneck a downsizing must avoid.

Question 691

A Java web service uses a mainstream Linux image, has ARM-compatible dependencies, and is mostly compute-bound. The team seeks lower EC2 cost per completed request. Which experiment?

Recompile only the application launcher for arm64 and assume every binary dependency will run on Graviton without compatibility testing.
Switch the whole fleet to Graviton solely because ARM list prices are lower, without testing JVM performance and native dependency compatibility.
Compare only the EC2 instance hourly rates and ignore whether the service completes requests more slowly on the new architecture.
Benchmark a compatible AWS Graviton instance family against the existing x86 fleet using representative throughput and latency.

Explanation

Correct Answer

Price-performance gains depend on actual workload, native libraries and architecture compatibility, not instance price alone.

Incorrect Answers

Graviton uses ARM64 instruction sets; native libraries and third-party binary compatibility must be validated alongside throughput.

Lower list price alone does not prove lower cost per completed request. Architecture compatibility and measured latency/throughput are critical to a Graviton evaluation.

A lower hourly price can be more expensive per successful request when throughput falls. Benchmark throughput, latency, resource use and compatibility together.

Question 692

An API fleet peaks at 200 instances during business hours but needs only 15 overnight. Which configuration prevents paying for 200 idle instances all night?

Use EC2 Auto Scaling policies with validated minimum capacity, scale-in behavior and demand metrics.
Set the ASG minimum and desired capacity permanently to the daytime peak of 200 even after the documented overnight load falls.
Individually stop 185 EC2 instances behind an Auto Scaling group while leaving that group desired capacity at 200, expecting the reduced server count to remain stable overnight.
Increase only the maximum Auto Scaling group size and assume that automatically reduces current desired capacity at night.

Explanation

Correct Answer

Dynamic scaling reduces capacity when load falls while preserving headroom for actual SLOs.

Incorrect Answers

Fixed maximum capacity as minimum prevents automatic scale-in and wastes EC2 hours during low traffic.

An Auto Scaling group replaces missing capacity to restore its desired count. Lower desired/minimum levels through suitable scaling policies or schedules, rather than stopping managed instances underneath the group.

Maximum capacity is a ceiling, not a scale-in instruction. Suitable policies, metric targets, minimums and cooldown/stabilization behavior govern downsizing.

Question 693

A development environment is needed only on weekdays from 09:00 to 17:00 and has predictable startup time. Which arrangement reduces unnecessary compute hours?

Keep the development EC2 fleet running 24x7 and rely on AWS Budgets email alerts to stop it automatically outside business hours.
Use scheduled scaling or stop/start automation with documented timezone, dependency and shutdown safeguards.
Leave the development instances always on and change only the EC2 instance type while ignoring the reliably idle evenings and weekends.
Stop every development service daily without testing database, storage and scheduled-job dependencies or expected startup time.

Explanation

Correct Answer

Predictable nonproduction duty cycles can be planned without running full capacity overnight.

Incorrect Answers

Budgets alerts alone do not replace a scheduled resource-lifecycle automation and will not stop the instance on a fixed timetable.

Rightsizing can change per-hour rates, but unused powered-on hours remain billed. Scheduling eligible stop/start removes idle compute hours with dependency safeguards.

Uncoordinated stoppage can disrupt active work or data services; a schedule must honor dependencies, time zones, persistence and a tested restart window.

Question 694

An internal webhook has fewer than 100 invocations per day and completes in seconds. Maintaining an always-on EC2 instance is its main cost. Which compute model is worth comparing?

Keep a dedicated always-on EC2 instance for the webhook, expecting fewer than 100 daily calls to fully utilize its provisioned compute.
Migrate the tiny webhook to provisioned Lambda concurrency that stays warm all month without comparing its additional idle charge.
Run the webhook as a continuously deployed one-task ECS/Fargate service that stays active every minute, assuming a short request handler is billed only when HTTP traffic arrives.
Use AWS Lambda with appropriate API/event integration and verify duration, memory, concurrency and cold-start requirements.

Explanation

Correct Answer

Event-driven billing can avoid paying for idle compute when invocations are sparse.

Incorrect Answers

A sparse short-duration event is often cheaper on request-based Lambda after evaluating execution and cold-start constraints.

Provisioned concurrency has ongoing capacity charges and may undermine savings for very infrequent invocations; basic on-demand Lambda can fit short sparse events.

A continuously running Fargate service is billed for task resource time even when idle. A Lambda handler can be billed per request/duration for sparse invocations.

Question 695

A small service team runs short-lived container jobs on ECS but does not want to own EC2 nodes, ASG patches or unused host capacity. Which launch model should it compare?

Launch large always-on ECS EC2 hosts for jobs that run only briefly each day, ignoring host utilization and patching overhead.
Use ECS on AWS Fargate and pay for task CPU/memory while accounting for minimum billing and workload patterns.
Choose a Spot-only EC2 container fleet without task retry or instance-health automation, assuming interruptible nodes eliminate all operational management.
Use ECS on Fargate and assume task CPU/memory reservations do not affect price if the containers use less in practice.

Explanation

Correct Answer

Fargate avoids EC2 host operations and aligns compute use with running task resource allocations.

Incorrect Answers

Fargate can remove host management for compatible tasks and charge task resource duration, though task sizing must still be evaluated.

Spot can reduce eligible EC2 compute price but does not remove host patching, capacity orchestration or reclaim handling. Fargate avoids the requested EC2 host-management work.

Fargate billing is based on configured task resources and running time, not only instantaneous observed CPU and memory usage. Rightsize task definitions.

Question 696

A developer operates an EC2 GUI analysis session that must retain open programs across overnight breaks but can tolerate an interruption to the instance. Which lifecycle feature may fit?

Use a normal EC2 stop and expect RAM contents and an open GUI session to resume unchanged on the next boot.
Take periodic EBS volume snapshots and assume that restoring them after normal EC2 shutdown also restores every live GUI process and in-memory application state.
Replace hibernation with an EC2 instance-store root volume and expect volatile instance-store data to preserve a stopped GUI session.
Use EC2 hibernation on a supported instance/AMI with sufficient encrypted EBS backing and valid lifecycle policy.

Explanation

Correct Answer

Hibernation preserves RAM state to disk for supported configurations whereas normal stop loses live application state.

Incorrect Answers

Stop/start preserves eligible EBS data but not running process memory; supported hibernation can preserve RAM state subject to its constraints.

EBS snapshots capture persisted block data, not automatically the guest RAM contents and running session state. EC2 hibernation deliberately saves RAM to encrypted EBS on supported configurations.

Instance-store data is ephemeral across stopping and hibernation does not support every root-volume configuration; use a supported hibernation/EBS setup.

Question 697

An RDS/Aurora database alternates between idle weekdays and intense month-end reports, but must keep relational semantics. What purchasing versus elasticity decision?

Purchase an inflexible three-year provisioned RDS commitment sized to the rarest monthly peak, assuming idle weeks cost nothing.
Compare Aurora Serverless v2 ACU-based capacity with provisioned RDS plus eligible reserved-instance discounts under measured demand.
Place RDS Proxy in front of a permanently oversized provisioned RDS instance and assume connection pooling scales the database CPU down to zero during idle weekdays.
Set an Aurora Serverless v2 maximum ACU below the load needed by month-end reports and treat serverless scaling as unlimited.

Explanation

Correct Answer

Serverless can match variable utilization while reserved provisioned database compute may be cheaper for a steady floor.

Incorrect Answers

A long fixed compute commitment can be underused when traffic is highly intermittent; elastic Aurora Serverless v2 may fit measured peaks.

RDS Proxy manages client/database connections; it does not transform provisioned instance compute billing into demand-based serverless ACUs. Evaluate actual provisioned utilization and eligible Aurora Serverless v2 characteristics.

Serverless v2 scales only within configured and supported ACU boundaries; a low maximum can throttle peak workload and undermine the expected elasticity.

Question 698

A fleet owner wants evidence-backed EC2 instance family and size recommendations using measured CPU, memory and network data, not guesses based on instance name. Which AWS tool?

Use the smallest EC2 family shown in the pricing catalog without examining real memory, network, EBS throughput or peak response times.
Review AWS Compute Optimizer recommendations alongside representative workload measurements and validation tests.
Use only the EC2 average CPU utilization to choose a smaller instance while ignoring memory exhaustion, EBS throughput and the infrequent month-end batch.
Apply the recommended downsizing directly to every production server with no representative load or rare-peak verification.

Explanation

Correct Answer

Compute Optimizer analyzes resource utilization to propose performance/cost tradeoffs; recommendations are candidates for testing.

Incorrect Answers

Compute Optimizer and workload telemetry are needed to compare resource headroom and SLO performance before rightsizing.

Average CPU is one capacity signal but not a proof of appropriate instance RAM, storage or network resources. Compute Optimizer and representative peak validation must consider the relevant bottlenecks.

Historical recommendations can miss unusual critical peaks; test the proposed size against service-level objectives before fleet-wide changes.

Question 699

Thousands of idempotent image-conversion jobs arrive in a queue each night; each can be retried if an instance disappears. Which Batch compute environment can lower cost?

Run AWS Batch only with a static On-Demand EC2 pool and claim it receives Spot interruption discounts without Spot capacity.
Select an AWS Batch Fargate On-Demand compute environment and assume that the word Fargate automatically makes those tasks Spot-priced regardless of capacity provider.
Use AWS Batch with Spot-capable compute capacity, suitable job retry policies and diversified instance selections.
Pin the entire Spot-backed AWS Batch fleet to one instance type and Availability Zone to guarantee jobs never wait for capacity.

Explanation

Correct Answer

Batch schedules queued jobs onto low-cost interruptible capacity when the workflow tolerates retry.

Incorrect Answers

Batch is an orchestration layer; Spot rates require selecting supported Spot-backed compute and designing retries.

AWS Batch Fargate and Fargate Spot are distinct capacity options with separate suitability and billing. Specify a supported Spot compute environment for interruptible discounts.

A single constrained Spot pool makes supply interruptions and insufficient-capacity errors more likely; diversify eligible pools and use retries/checkpoints.

Question 700

A product team must receive a notification when its AWS monthly forecast crosses a spending threshold, before the final invoice arrives. Which service?

Configure AWS Budgets with a cost or usage budget and appropriate actual/forecast alert recipients.
Create a cost allocation tag but no budget thresholds, subscriber addresses or notification actions, and expect automatic forecasts.
Create only a Cost Explorer historical report and assume it automatically emails a threshold alert whenever a future monthly forecast crosses the budget.
Create an AWS Budget threshold but leave notification subscribers and actions empty, assuming alert delivery happens for every account by default.

Explanation

Correct Answer

Budgets compares observed/forecast spend to thresholds and sends alerts; a cost report alone is retrospective.

Incorrect Answers

Cost allocation labels are accounting dimensions; AWS Budgets needs an actual configured budget threshold and notification recipient.

Historical cost analysis is not a budget notification subscription. AWS Budgets needs a defined actual/forecast threshold and configured notification recipient.

A calculated threshold alone does not deliver the email or SNS alert. Configure notification recipients or supported actions and verify delivery.

Pass Your Amazon AWS Certified Solutions Architect - Associate SAA-C03 Exam Easy!

Question 690

Explanation

Correct Answer

Incorrect Answers

Question 691

Explanation

Correct Answer

Incorrect Answers

Question 692

Explanation

Correct Answer

Incorrect Answers

Question 693

Explanation

Correct Answer

Incorrect Answers

Question 694

Explanation

Correct Answer

Incorrect Answers

Question 695

Explanation

Correct Answer

Incorrect Answers

Question 696

Explanation

Correct Answer

Incorrect Answers

Question 697

Explanation

Correct Answer

Incorrect Answers

Question 698

Explanation

Correct Answer

Incorrect Answers

Question 699

Explanation

Correct Answer

Incorrect Answers

Question 700

Explanation

Correct Answer

Incorrect Answers

AWS Certified Solutions Architect - Associate SAA-C03 Practice Tests