Amazon AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 73

Question 720

A team adopts a Compute Optimizer downsizing recommendation without testing a rare end-of-quarter workload and breaches its latency SLO. What was wrong?

A high-level EC2 CPU chart accurately represents every instance’s RAM and I/O demands for all time, even if peak workloads are missing.
Compute Optimizer guarantees that a historical utilization recommendation has already passed every future business-specific latency SLO.
Change the EC2 instance family on the recommendation without rechecking operating-system, CPU-architecture and native-library compatibility with the end-of-quarter batch.
Historical recommendation input did not establish performance under the unrepresented peak workload.

Explanation

Correct Answer

Recommendations should be validated against the target workload and memory/network/IO headroom.

Incorrect Answers

CPU alone is insufficient: representative memory, disk, network and seasonal load can make a smaller instance unsuitable.

Recommendations derive from observed metrics and assumptions, not unseen rare workloads; representative peak testing remains necessary.

Even if the historical utilization is representative, a family or architecture change needs application compatibility checks. The unobserved peak memory/I/O demand in this case also requires load testing.

Question 721

AWS Batch launches jobs on Spot, but interrupted jobs are marked failed permanently with no retry allowance. What was missing?

AWS Batch automatically retries every interrupted job infinitely even when job definitions forbid retry or checkpoints are unavailable.
The application/job definition and Batch retry strategy did not account for expected Spot interruptions.
Set the AWS Batch job queue priority higher than competing workloads and assume a permanently failed Spot-interrupted job will be automatically resubmitted without retry rules.
Increasing a job queue priority alone guarantees AWS Batch will re-run every failed interrupted job despite an explicit zero-retry strategy.

Explanation

Correct Answer

Define checkpoint and retry conditions rather than assuming Batch automatically guarantees completion on reclaimed capacity.

Incorrect Answers

Interruption-safe jobs need appropriate retry strategies and durable state; Batch does not guarantee infinite idempotent progress.

Scheduling priority influences which pending jobs run; it does not change the retry attempts or checkpoint recovery for an already failed job.

Queue priority affects scheduling order, not the number of retries permitted by a job definition or submission.

Question 722

A team creates an AWS Budget with no email or SNS subscribers and expects immediate notifications at 80%. Nothing is received. Why?

AWS Budgets notifies account operators on every forecast crossing without any notification subscriber or alarm configuration.
Create an Amazon SNS topic in the account but never subscribe an email endpoint or connect the topic to this budget’s notification, assuming the topic’s mere existence sends threshold alerts.
A budget threshold without a configured notification action/recipient does not deliver alerts.
AWS Budgets ignores its own alert recipients whenever the configured threshold is less than one hundred percent.

Explanation

Correct Answer

Check the budget’s alert definitions, notification subscriptions and monitoring refresh timing.

Incorrect Answers

A budget or threshold alone does not guarantee delivery; it needs specified alert recipients or supported notification actions.

An SNS topic needs subscriptions and the AWS Budget must reference an appropriate alert destination; neither an unrelated topic nor a threshold without recipients delivers notices.

Alert thresholds can be below 100 percent; the missing notification recipient is the operative problem, not the 80 percent threshold.

Question 723

An engineer adds a new resource tag today then expects Cost Explorer to allocate past six months of already billed usage under that tag. Why not?

Activating a cost allocation tag today automatically backfills six months of missing tag metadata across all previously billed usage.
Apply a newly enabled resource tag only to today’s replacement instances and assume Cost Explorer can infer that tag for terminated instances from the six months before it existed.
Cost Explorer uses tags as a billing discount, so historical charges disappear instead of being grouped when the tag is activated.
Newly activated cost allocation tags do not automatically retroactively label historical usage records that lacked the tag.

Explanation

Correct Answer

Use the relevant historical allocation method and ensure consistent tagging going forward.

Incorrect Answers

Cost allocation tags do not retroactively populate missing historical usage records, so old periods can remain unattributed.

Tagging new resources cannot reconstruct missing historical tag values on terminated instances. Reporting from previous months needs data that was actually captured or a separately supported allocation method.

Tags support attribution, not pricing discounts; they cannot reconstruct six months of previously untagged resource usage.

Question 724

An internal cost report compares only unblended upfront RI payment in purchase month and reports near-zero costs for other months. What accounting error?

The report distributes the RI upfront commitment cost across the covered future months rather than assigning it all to the purchase month.
The report is not amortizing commitment costs across covered usage periods, so monthly allocation is distorted.
Allocating the full upfront RI purchase amount only to the purchase month accurately represents effective covered-use cost in later months.
The organization should remove RI and Savings Plan columns from its CUR/Data Exports because amortized costs are already contained in unblended on-demand line items.

Explanation

Correct Answer

Use appropriate amortized/net amortized fields in detailed billing datasets.

Incorrect Answers

That is the correct amortization approach, not the defect described. The defective report fails to distribute upfront commitment cost into covered future usage periods.

Cash payment timing differs from amortized monthly cost; chargeback should allocate commitment spend over the benefit period and actual covered usage.

Detailed amortization and covered-usage fields are needed to attribute commitments; ordinary unblended cash amounts do not substitute for amortized allocation.

Question 725

An ECS cluster keeps scaling out hosts despite CPU utilization below 20%; task CPU reservations are near full capacity. Why?

ECS task placement can pack unlimited CPU and memory reservations onto a node if reported process utilization is near zero.
ECS placement considers only current observed host CPU percentages, ignoring configured task CPU and memory reservations.
Scheduler placement is constrained by declared task reservations, not merely actual observed host CPU use.
Increasing the ASG maximum host count automatically improves ECS task bin packing without revisiting oversize task CPU reservations.

Explanation

Correct Answer

Review task definition requests and placement constraints to avoid artificial resource exhaustion.

Incorrect Answers

Scheduler reservations constrain placement even when current usage is low; overreserved tasks can waste EC2 node hours.

Schedulers must honor declared reservations; high requests can block placements even when runtime CPU utilization is low.

A higher ceiling permits more hosts but does not improve the inefficient resource request fit that created unnecessary scale-out.

Question 726

An organization chooses Dedicated Instances for socket-based BYOL but cannot show host-level socket inventory needed for auditing. What distinction matters?

Dedicated Hosts guarantee compliance with every software vendor contract merely by being purchased, regardless of socket limits and license rules.
Dedicated Instances expose the same persistent physical host identity and placement controls as Dedicated Hosts for a socket-counted license.
A shared-tenancy EC2 instance proves ownership of the underlying physical CPU sockets as long as the guest operating system reports vCPUs.
Dedicated Instances isolate tenancy but do not provide the same dedicated physical host identity/control as Dedicated Hosts.

Explanation

Correct Answer

Choose the licensing-compatible tenancy based on vendor terms and available host accounting.

Incorrect Answers

License compliance still requires checking actual vendor entitlement, sockets/cores, allocation and supported AWS deployment conditions.

Dedicated Instances use isolated tenancy but do not provide the required physical host control/reporting semantics of Dedicated Hosts.

Guest vCPU counts do not establish a dedicated physical host or socket licensing inventory; Dedicated Hosts are the relevant tenancy model.

Question 727

A developer chooses the smallest Lambda memory value as always cheapest, but function CPU bottleneck stretches duration sharply. What is missing?

Lambda bills for allocated memory multiplied by run duration, so slower low-memory executions can cost more than faster high-memory ones.
Lambda memory doubling necessarily doubles the full request cost even if it reduces the execution time by a factor of ten.
Lambda duration is independent of its configured memory, so every memory upgrade increases costs proportionally without any runtime gain.
The smallest-memory function always incurs the fewest billed GB-seconds even when it runs far longer due to constrained CPU.

Explanation

Correct Answer

Optimize the product of configured resources and measured time, not RAM size alone.

Incorrect Answers

Billed compute is based on configured memory multiplied by execution duration; faster high-memory runs can sometimes cost less.

Lambda allocates more CPU with higher configured memory; shorter duration can offset a higher rate, requiring benchmarks of both dimensions.

GB-second charges multiply configured memory and billed duration; low memory can prolong runtime enough to increase total charges.

Question 728

A Spot fleet uses only one instance type/AZ and reports no capacity available during a demand spike. What design limit did it impose?

Spot interruption recommendations are a guarantee that an instance continues running indefinitely once a rebalance notification is handled.
A single constrained Spot instance-type and Availability Zone selection guarantees lower placement delays by preventing capacity-pool changes.
Increasing the Spot maximum price automatically creates unused instances in the sole selected pool when no actual spare capacity exists.
Overconstraining Spot capacity pools reduces the alternatives the allocator can use when spare supply shifts.

Explanation

Correct Answer

Diversify compatible instance types and AZs as workload requirements permit.

Incorrect Answers

Rebalance and interruption signals help proactive recovery, not an availability guarantee for spare capacity.

Constraining supply removes alternatives when the chosen pool is depleted; diversification improves the chance of available spare capacity.

Price tolerance cannot manufacture unavailable capacity. Broader eligible instance types, sizes and Availability Zones improve resilience to shortages.

Question 729

A DynamoDB event table handles occasional sudden multi-million-request campaigns but has near-zero baseline traffic. The team dislikes paying provisioned capacity during idle weeks. Which capacity model?

Provision the DynamoDB table permanently for the maximum campaign request rate even though it remains almost unused most days.
Keep a small provisioned DynamoDB RCU/WCU allocation with target-tracking auto scaling and assume an abrupt campaign can always jump immediately to any peak without capacity or scaling-history considerations.
Evaluate DynamoDB on-demand request pricing and documented scaling behavior for the burst pattern.
Buy provisioned capacity sized to a rare peak and assume DynamoDB credits every unused read/write capacity unit back during the following idle weeks.

Explanation

Correct Answer

On-demand pricing tracks read/write requests without a constant provisioned RCU/WCU bill, subject to throughput limits.

Incorrect Answers

A spiky workload with a near-zero baseline can waste provisioned capacity; on-demand pricing should be evaluated with peak scaling considerations.

Target-tracking provisioned auto scaling responds to measured utilization and can have adjustment delays and limits. It does not guarantee unthrottled, instant scaling when an unpredictable campaign jumps far beyond the table’s provisioned capacity.

Provisioned capacity is billed for capacity allocated over time, even when little traffic arrives. Idle periods do not create a per-request refund for unused throughput; on-demand pricing should be evaluated for this sharply variable traffic profile.

Question 730

A payment-lookup DynamoDB workload runs close to an established baseline around the clock, with predictable moderate peaks. Which mode may lower request cost compared with on-demand?

Choose DynamoDB on-demand mode without comparing its per-request rate to reliably utilized provisioned capacity over the same hourly period.
Use provisioned RCU/WCU capacity permanently set to the traffic trough and leave auto scaling disabled, assuming DynamoDB will absorb every predictable daytime peak without throttling.
Compare provisioned DynamoDB capacity with auto scaling and reserved capacity where eligible against actual request totals.
Switch to DynamoDB Standard-IA solely to reduce read/write request charges for the high-throughput payment workload.

Explanation

Correct Answer

Predictable usage can benefit from provisioned RCU/WCU economics if utilization remains high enough.

Incorrect Answers

When demand is stable, well-used provisioned capacity and supported auto scaling can be cheaper; neither model is universally best.

The provisioned option can be economical for predictable demand, but static capacity sized to a trough may throttle at a higher known peak. Auto scaling and a sufficient minimum are needed to match the measured hourly demand.

Standard-IA decreases storage cost but increases the price of requests relative to Standard. Table class is not the same as provisioned-versus-on-demand throughput mode and is not automatically cheaper for request-heavy payments.

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

Question 720

Explanation

Correct Answer

Incorrect Answers

Question 721

Explanation

Correct Answer

Incorrect Answers

Question 722

Explanation

Correct Answer

Incorrect Answers

Question 723

Explanation

Correct Answer

Incorrect Answers

Question 724

Explanation

Correct Answer

Incorrect Answers

Question 725

Explanation

Correct Answer

Incorrect Answers

Question 726

Explanation

Correct Answer

Incorrect Answers

Question 727

Explanation

Correct Answer

Incorrect Answers

Question 728

Explanation

Correct Answer

Incorrect Answers

Question 729

Explanation

Correct Answer

Incorrect Answers

Question 730

Explanation

Correct Answer

Incorrect Answers

AWS Certified Solutions Architect - Associate SAA-C03 Practice Tests