Amazon AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 51

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.

Question 501

A distributed HPC solver requires very low inter-instance latency and high throughput among tightly coupled EC2 nodes. Which placement strategy supports proximity?

Cluster placement group in a suitable AZ with compatible instance/network types.
Spread placement group explicitly isolates each node on distinct underlying hardware.
Partition placement group is solely for low-latency dense communication on the same hardware.
Distribute tightly coupled workers across multiple distant Availability Zones and assume their cross-AZ network RTT matches colocated cluster placement within one AZ.

Explanation

Correct Answer

Cluster placement groups co-locate EC2 capacity for low-latency node-to-node communication.

Incorrect Answers

Spread favors fault separation, not tight network proximity.

Partition distributes fault domains for replicated systems.

Cluster placement in a supported AZ reduces network distance for tightly coupled nodes; spanning AZs adds path latency and loses the particular placement benefit.

Question 502

Seven critical independent domain controllers should avoid sharing underlying hardware faults in one AZ. Which placement option matches that small count?

Cluster placement group to pack all controllers together.
Spread placement group subject to the per-AZ running-instance limit and capacity.
One shared EC2 instance running seven controllers as containers.
Place all seven domain controllers in a single cluster placement group and assume concentrated physical placement minimizes correlated underlying host failures.

Explanation

Correct Answer

Spread distributes eligible instances across distinct hardware to reduce correlated host failures.

Incorrect Answers

Cluster proximity increases potential shared hardware impact.

That concentrates failures rather than spreading them.

Cluster placement is optimized for low latency rather than strong hardware-failure separation. Spread placement targets distinct hardware for a small supported number of instances.

Question 503

A distributed Cassandra cluster wants its EC2 replicas grouped into separately identifiable hardware failure partitions across AZs. Which placement group model is suited?

Partition placement groups with replicas spread across separate partitions under supported limits.
Cluster placement only, placing all replicas together for shortest RTT.
Spread placement for hundreds of instances in a single AZ, ignoring seven-running-instance limit.
Use a spread placement group for several hundred Cassandra nodes per AZ and assume the group has no per-AZ placement limit.

Explanation

Correct Answer

Partition placement helps distributed databases place replicated data across distinct hardware partitions.

Incorrect Answers

Cluster proximity can reduce fault isolation.

Spread has restrictive per-AZ instance limits.

Spread groups have per-AZ limits and are meant for a small number of critical instances, not scalable fault partitions for a large distributed dataset.

Question 504

Thousands of compute-heavy container jobs arrive irregularly and can wait for scheduling. The company wants managed job priorities and elastic compute allocation. Which AWS service fits?

AWS Batch with appropriate job definitions, queues, compute environments and resource scaling.
An Application Load Balancer as the sole queued job scheduler.
One fixed EC2 instance with no queue to process all jobs.
Use ECS Service Auto Scaling with one continuously running fixed service deployment and assume it provides AWS Batch-style queued priorities and independent job retry scheduling automatically.

Explanation

Correct Answer

Batch schedules queued jobs onto elastic capacity without requiring an interactive service endpoint.

Incorrect Answers

ALB routes live HTTP requests rather than creating compute batches.

A single host becomes a throughput bottleneck and resource-management burden.

ECS can run workers, but a regular service Auto Scaling policy is not itself a managed multi-queue batch scheduler with job definitions, retries and priorities. AWS Batch matches the stated job workflow.

Question 505

A worker fleet processes jobs from SQS but CPU is low while queue backlog keeps increasing from blocked downstream I/O. What better Auto Scaling signal matches work pressure?

A suitably calculated queue backlog-per-instance or oldest-message age target with worker performance measurements.
Use raw queue length as a directly proportionate target tracking metric without normalizing by workers.
Keep CPU as the only scaling signal despite known I/O stalls.
Configure target tracking on the SQS queue ApproximateNumberOfMessagesVisible count as if it were already normalized per active worker without a custom backlog-per-instance metric.

Explanation

Correct Answer

Per-worker queue demand scales meaningfully with instance count despite low CPU.

Incorrect Answers

Raw length may not scale inversely with fleet capacity.

CPU cannot observe all job service-time bottlenecks.

Raw queue length is not a per-worker target; workers and processing time determine actual backlog pressure. Configure a suitable normalized or age-based metric with demand measurements.

Question 506

An application is built in portable Java and may move from x86 EC2 to AWS Graviton. Which evidence should determine the adoption decision?

Benchmark functionally equivalent arm64 deployments for correctness, dependency support, performance and cost.
Assume every x86 native library runs on arm64 without rebuilding.
Choose Graviton solely from a marketing price statement without workload profiling.
Change the EC2 VPC subnet route table to use a shorter network path and assume network routing translates x86 native libraries into compatible arm64 instructions.

Explanation

Correct Answer

Graviton may improve price-performance but architecture/native dependency compatibility must be measured.

Incorrect Answers

Machine architecture and binary compatibility differ.

Actual performance and support vary by application.

Network routing can change reachability/latency, but cannot solve instruction-set or native dependency compatibility. Build/test arm64 artifacts and benchmark the real workload.

Question 507

A developer workstation runs memory-intensive tools that take hours to reopen. It must stop overnight to save compute cost yet resume process RAM state. Which EC2 feature is relevant?

Instance hibernation if instance type/AMI, EBS root space and feature prerequisites are satisfied.
Ordinary stop/start always preserves RAM and running processes.
Instance termination followed by new launch with no saved image.
Enable EC2 user-data scripts to relaunch processes after ordinary stop/start and assume newly launched processes preserve their exact in-memory heap and working state.

Explanation

Correct Answer

Hibernation saves memory contents to EBS so processes resume after start rather than cold boot.

Incorrect Answers

Normal stop ends memory state; hibernation is different.

Termination discards runtime process state.

Startup scripts can restart applications but cannot recreate arbitrary prior RAM/process state. Supported EC2 hibernation explicitly saves memory contents when its prerequisites are met.

Question 508

A software build compiles hundreds of independent modules on a 32-vCPU EC2 host, but monitoring shows only one vCPU busy during a long serial link phase. Doubling to a 64-vCPU instance barely reduces total build time. What performance constraint explains the result?

The EBS gp3 volume must be saturated because a host with many idle vCPUs cannot have a serial CPU bottleneck.
The compiler must be using the wrong EC2 Region, because moving Regions automatically enables parallel compilation for every serial link task.
The build has a serial critical-path phase that extra cores cannot accelerate without algorithm or build-pipeline changes; profile that phase before purchasing more vCPUs.
Increasing the number of allocated vCPUs guarantees the serial linker divides one dependency chain equally among all cores without modifying its execution plan.

Explanation

Correct Answer

A serial dependency or single-threaded linker can dominate end-to-end duration; Amdahl-style limits mean more idle cores do little. Optimize the build graph, parallelizable work or single-core performance rather than buying unused core count.

Incorrect Answers

Storage can affect a build, but the observed single fully busy core with many idle cores points to a serial compute phase. Profile CPU, storage and dependency scheduling before assuming an EBS saturation.

Region location alone does not parallelize a serialized algorithm or link step. The build graph and compiler/linker implementation govern how many cores can execute useful concurrent work.

Extra vCPUs help only tasks that can execute concurrently. A serial critical path does not become parallel merely because the EC2 instance exposes more logical CPUs.

Question 509

An EC2-hosted analytics daemon steadily increases its resident memory over a day. The input size stays constant; process heap profiles show old allocations retained after requests complete, and host swapping begins late in the day. What should the team investigate first?

Increase the instance vCPU count because CPU capacity automatically releases retained heap objects after each request.
Investigate application memory retention or a leak using allocation/heap profiles and repair its lifetime management before relying on more RAM.
Reduce the number of EBS snapshots because snapshot count is the cause of heap allocations remaining live in the process.
Set Route 53 health-check thresholds lower so DNS policy removes unused heap objects inside the running process.

Explanation

Correct Answer

The progressive growth and heap profiles distinguish a retention defect from a simply large but stable working set. Investigate reference lifetime, unbounded caches and allocation paths, then size memory with a measured steady-state baseline.

Incorrect Answers

Additional CPUs do not force a process to free references or objects it has retained. Steady heap growth at a stable workload points to memory-lifetime management, which CPU-only resizing does not address.

Snapshots are storage recovery objects. They do not control live process allocations or explain why old memory remains reachable after requests finish.

DNS health checks influence client routing, not the program allocator. They cannot fix a leak inside an EC2 host process.

Question 510

A T3 application server is configured in Unlimited mode. It holds CPU utilization above its baseline for many consecutive hours without the throttling seen in Standard mode, but the bill now contains CPU credit charges. Why?

Unlimited mode makes surplus CPU credits permanently free once the original accrued balance reaches zero.
Above-baseline sustained CPU can consume surplus credits that incur charges when usage exceeds the applicable credit earning and grace economics.
EBS gp3 IOPS are billed under the name CPU credits whenever an instance opens a large file.
An ASG health check increases the CPU credit balance on every instance without limits or price impact.

Explanation

Correct Answer

With Unlimited, usage above baseline can draw surplus credits and incur the documented additional charge under the applicable averaging rules. Compare sustained workload price/performance with a nonburstable instance before assuming the T-family is the economical fit.

Incorrect Answers

The Unlimited mode permits prolonged bursts, but its billing treatment is not identical to free baseline entitlement. Sustained excess CPU can be charged.

EBS I/O charges and EC2 CPU-credit charges represent different metering dimensions. Disk IOPS does not relabel itself as compute surplus credits.

Health checks can replace unhealthy fleet members, not mint free credits. The Unlimited-mode surplus accounting is driven by the instance CPU use and credit rules.

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

Question 500

Explanation

Correct Answer

Incorrect Answers

Question 501

Explanation

Correct Answer

Incorrect Answers

Question 502

Explanation

Correct Answer

Incorrect Answers

Question 503

Explanation

Correct Answer

Incorrect Answers

Question 504

Explanation

Correct Answer

Incorrect Answers

Question 505

Explanation

Correct Answer

Incorrect Answers

Question 506

Explanation

Correct Answer

Incorrect Answers

Question 507

Explanation

Correct Answer

Incorrect Answers

Question 508

Explanation

Correct Answer

Incorrect Answers

Question 509

Explanation

Correct Answer

Incorrect Answers

Question 510

Explanation

Correct Answer

Incorrect Answers

AWS Certified Solutions Architect - Associate SAA-C03 Practice Tests