Amazon AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 50

Question 490

An ALB-backed web tier sees variable traffic while CPU remains low because most work waits on downstream I/O. Which autoscaling metric can directly reflect request load per healthy target?

Target tracking on ALBRequestCountPerTarget with an appropriate request-per-target target value.
A fixed total ALB RequestCount as the only target tracking metric.
Set the ASG scale-out target to instance disk capacity remaining even though the observed pressure is HTTP requests waiting on downstream I/O.
Use only Route 53 DNS lookup counts for the app hostname and assume cached DNS responses exactly equal HTTP requests served by each healthy ALB target.

Explanation

Correct Answer

Per-target request demand responds to changes in fleet size and can track I/O-bound web traffic better than CPU.

Incorrect Answers

Total requests do not decrease per target when capacity doubles.

Available root filesystem GiB does not directly measure incoming ALB requests per serving target. Choose a scaling signal that correlates with actual work, such as ALB request count per target.

DNS resolvers cache lookups and may reuse an answer for many HTTP requests; DNS question counts do not directly represent load per ALB target.

Question 491

A media site sees large weekday traffic peaks that begin faster than EC2 hosts can initialize. It has months of repeatable request history. Which Auto Scaling technique can warm capacity in advance?

Keep ASG minimum at zero during every peak and wait for CPU alarms.
Predictive scaling from historical cyclical load, combined with reactive scaling where appropriate.
Apply Auto Scaling step scaling only after the traffic spike has already overloaded the tier and assume it always warms the new EC2 instances before repeatable peaks.
Use a one-time manual instance launch without forecasting or maintenance.

Explanation

Correct Answer

Forecast-driven scale-out can prepare instances before repeatable daily/weekly peaks.

Incorrect Answers

Reactive cold starts may miss the SLO.

Reactive step scaling depends on current metrics, so launch/boot latency can leave a known predictable peak underprovisioned. Predictive scaling can pre-launch forecast capacity.

A manual action does not track recurring demand patterns.

Question 492

A nonproduction processing fleet is needed exactly between 8am and 5pm on weekdays. No historical forecasting is required. What is a simple cost-aware scaling mechanism?

Maintain peak desired capacity overnight regardless of use.
Use Predictive Scaling exclusively even though the calendar is fixed and known.
Use EC2 Instance Scheduler tags without defining or deploying any actual recurring schedule, assuming tags alone automatically stop/start or scale an ASG to desired weekday capacity.
EC2 Auto Scaling scheduled actions to increase desired capacity before work and reduce it after hours.

Explanation

Correct Answer

Known fixed schedules can adjust capacity without waiting for demand signals.

Incorrect Answers

It pays for unnecessary idle compute.

Complex forecasting is unnecessary for a deterministic window.

Resource tags describe intent but do not implement scaling events. Auto Scaling scheduled actions can apply deterministic desired capacity changes at configured times.

Question 493

Thousands of independent rendering tasks save checkpoints and can restart after a short interruption. The owner wants lower compute cost and accepts reclaim risk. Which purchase model fits?

Use EC2 Spot capacity with an interruption-aware queue/checkpoint strategy.
Use Spot as a guarantee every instance runs uninterrupted until the deadline.
Use Dedicated Hosts exclusively to lower all transient compute rates.
Purchase only On-Demand capacity for every fully checkpointable renderer because Spot interruption notices are always unrecoverable and checkpoint restarts are unsupported.

Explanation

Correct Answer

Spot discounts unused capacity but instances can be interrupted and jobs must recover safely.

Incorrect Answers

Spot can be reclaimed with short notice.

Dedicated tenancy often costs more and is not required here.

The scenario explicitly tolerates interruption and restart. Spot capacity can lower price when the queue and checkpoints safely handle interruption events, unlike insisting all work requires guaranteed On-Demand.

Question 494

A low-latency banking API cannot tolerate interruptions and its short-term workload is uncertain. Which EC2 pricing choice avoids Spot interruptions and long-term commitment?

Spot capacity for every sole production instance with no interruption recovery.
A three-year commitment for unmeasured transient demand as the only option.
On-Demand Instances with suitable HA capacity and monitoring.
Purchase long-term Compute Savings Plans based on an assumed capacity baseline that has not been measured, expecting they eliminate every future On-Demand commitment risk.

Explanation

Correct Answer

On-Demand provides flexible capacity without Spot reclamation or commitment-based discount assumptions.

Incorrect Answers

Spot reclamation violates availability requirements.

Long commitments are not justified before utilization is known.

Savings Plans create a committed hourly spend and need a reasonable utilization forecast; unpredictable short-term demand may favor On-Demand initially.

Question 495

An ECS service scales to four tasks at 30% CPU but has high API request counts and latency due to I/O waiting. Which service scaling technique is relevant?

Increase task count only after editing every task’s image name manually.
Set ECS service desiredCount to one permanently and increase EBS gp3 throughput, assuming disk bandwidth is the same metric as API requests waiting on downstream services.
Use only ECS service average CPU utilization for autoscaling even though the service is demonstrably I/O-wait bound at low CPU and request latency is growing.
Application Auto Scaling target tracking for a workload-appropriate ECS service utilization or request-related metric.

Explanation

Correct Answer

ECS task count should respond to measured bottleneck demand rather than CPU alone.

Incorrect Answers

Service autoscaling can adjust desired count automatically.

Storage tuning cannot provide more application tasks for a service under request pressure. Scale task capacity based on meaningful ECS service demand metrics after verifying downstream limits.

CPU can remain low while tasks are saturated on downstream operations. Use a request/queue/latency-based relevant metric or redesign I/O pressure rather than CPU-only scaling.

Question 496

An EKS service experiences request surges. New pods can handle load but the existing nodes lack enough schedulable CPU when HPA increases replica count. What complementary scaling mechanism is required?

Node provisioning or cluster autoscaling in addition to pod-level HPA, with resource requests/limits configured.
HPA automatically creates EC2 worker nodes in every unsupported configuration.
Stop registering pods with Kubernetes Services so they gain CPU.
Raise the HPA replica limit without adding worker node provisioning, assuming a high HPA cap itself supplies new schedulable node CPU.

Explanation

Correct Answer

HPA creates desired pods; compute nodes must actually have space to schedule them.

Incorrect Answers

Node lifecycle has separate controllers/capacity providers.

Service discovery does not create capacity.

HPA scales pod replica requests but cannot add node compute resources. Cluster Autoscaler/Karpenter or compatible node provisioning must supply capacity when pods are pending.

Question 497

A Lambda function is CPU-bound and completes slowly at its current low memory setting. What AWS Lambda resource knob should be benchmarked for lower duration?

Only extend timeout without addressing computational throughput.
Increase the Lambda timeout to 15 minutes and assume longer permitted execution automatically gives the function additional CPU cores for CPU-bound work.
Raise the function memory allocation in measured steps because CPU entitlement increases with memory.
Assume CPU is configured independently of memory for ordinary Lambda functions.

Explanation

Correct Answer

More memory can yield more CPU and shorter runtime, sometimes improving both latency and cost.

Incorrect Answers

A timeout allows more time but does not speed execution.

Timeout constrains maximum allowed duration, not CPU allocation. Lambda CPU entitlement scales with configured memory, so measure runtime and cost under larger memory settings.

Lambda’s normal resource allocation ties CPU share to memory.

Question 498

A Lambda handler writes to an external database that supports at most 50 concurrent connections. Traffic can exceed 400 simultaneous invokes. Which control protects the backend?

Increase Lambda reserved concurrency to the account maximum with no DB mitigation.
Enable Lambda provisioned concurrency for 400 warm environments while leaving unthrottled database connections at 50, assuming warm starts enforce the database connection limit.
Limit reserved concurrency appropriately and consider connection pooling/queueing.
Use Lambda function timeout alone to limit database connection concurrency, assuming short timeouts prevent hundreds of simultaneously open database sessions.

Explanation

Correct Answer

Reserved concurrency caps simultaneous function execution and therefore can bound downstream pressure.

Incorrect Answers

That can amplify connection exhaustion.

Provisioned concurrency helps initialization latency but does not directly cap backend connection count; reserved concurrency, pooling or queue design must protect the DB.

A shorter timeout bounds execution lifetime but does not directly limit the number of concurrent invocations starting database sessions. Use reserved concurrency or queueing.

Question 499

A user-facing Lambda function has intermittent first-invoke initialization delays. The business requires predictable p99 latency for a known baseline of concurrent requests. What setting should be evaluated?

Reserved concurrency alone guarantees every environment was warm before invocation.
A longer Lambda timeout removes initialization time from latency.
Configure an EventBridge scheduled warm-up invocation once each morning and assume it guarantees all demand-time concurrency slots remain initialized for the entire day.
Provisioned concurrency on the invoked published version or alias, sized to baseline demand.

Explanation

Correct Answer

Provisioned concurrency pre-initializes execution environments to reduce cold-start variability.

Incorrect Answers

Reserved concurrency caps/guarantees capacity but does not pre-initialize workers.

Timeout is a maximum duration, not cold-start removal.

Occasional warm-up calls cannot guarantee enough ready environments across all provisioned traffic bursts. Provisioned concurrency is intended for defined baseline warm capacity.

Question 500

A Fargate task extracts and sorts huge archives using temporary local files that exceed the default task scratch capacity. What resource should be configured?

Increase supported Fargate ephemeral storage in the task definition and monitor disk utilization.
Raise task CPU but leave the same full ephemeral disk.
Change SQS queue visibility timeout without expanding task storage.
Allocate additional task memory while keeping the Fargate ephemeral storage size below the archive’s temporary disk requirement, assuming memory alone expands local scratch disk.

Explanation

Correct Answer

Fargate allocates scratch storage separately from CPU and memory.

Incorrect Answers

CPU cannot add local storage space.

Queue visibility does not affect scratch capacity.

Task memory and Fargate ephemeral storage have separate settings. A disk-capacity failure needs a supported ephemeral-storage increase or streaming alternative.

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

Question 490

Explanation

Correct Answer

Incorrect Answers

Question 491

Explanation

Correct Answer

Incorrect Answers

Question 492

Explanation

Correct Answer

Incorrect Answers

Question 493

Explanation

Correct Answer

Incorrect Answers

Question 494

Explanation

Correct Answer

Incorrect Answers

Question 495

Explanation

Correct Answer

Incorrect Answers

Question 496

Explanation

Correct Answer

Incorrect Answers

Question 497

Explanation

Correct Answer

Incorrect Answers

Question 498

Explanation

Correct Answer

Incorrect Answers

Question 499

Explanation

Correct Answer

Incorrect Answers

Question 500

Explanation

Correct Answer

Incorrect Answers

AWS Certified Solutions Architect - Associate SAA-C03 Practice Tests