SOA-C02 Cost and Performance Optimization Practice Question
A company is running a high-performance computing (HPC) workload on EC2. The workload is time-sensitive and runs for 2 hours every night. The company wants to minimize costs. Which THREE options should they consider? (Choose THREE.)
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Use an EFS or S3 as shared storage instead of EBS volumes.
Options B, C, and E are correct. B is correct because using EFS or S3 as shared storage reduces costs compared to attaching individual EBS volumes to each instance. C is correct because using smaller instance types in parallel can be more cost-effective for HPC workloads that can be parallelized. E is correct because Spot Instances offer significant discounts for fault-tolerant workloads like HPC. Option A is incorrect because Reserved Instances require a 1-3 year commitment and would not be cost-effective for a 2-hour nightly job. Option D is incorrect because Dedicated Instances are more expensive and not necessary for this workload.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Purchase Reserved Instances for the nightly run.
Why it's wrong here
Reserved Instances require a 1- or 3-year capacity commitment, so you'd pay for full-time compute even though the nightly HPC run only consumes a few hours per day. That mismatch makes RIs virtually useless for a short, predictable batch job; the per-hour cost savings do not offset the idle hours you are billed for. For a non-continuous workload, you should instead use On-Demand with scheduling or Spot, not RIs.
- ✓
Use an EFS or S3 as shared storage instead of EBS volumes.
Why this is correct
Using a shared file system like Amazon EFS or object storage like Amazon S3 lets all HPC compute nodes access the same dataset without each node needing its own EBS volume copy. EBS volumes are per-instance block storage charged by GB-month plus IOPS, so duplicating data across hundreds of instances multiplies storage cost and also creates consistency problems when output is written locally. EFS automatically scales and is designed for shared POSIX access, while S3 is ideal for input/output datasets and checkpoint artifacts—both eliminate redundant EBS allocations.
- ✓
Use smaller instance types and distribute the workload.
Why this is correct
Choosing smaller instance types and spreading the compute across more nodes can lower the effective cost per vCPU-hour when the HPC workload is embarrassingly parallel or can be partitioned across a distributed cluster. Smaller instances often fall into lower pricing tiers per vCPU than a single largest instance, especially when using a mix of general-purpose or compute-optimized families. However, you must ensure the workload is network-bound and tolerates inter-node communication, because tightly coupled MPI jobs may see worse performance if the instance network fabric is not chosen carefully.
- ✗
Use Dedicated Instances for performance isolation.
Why it's wrong here
Dedicated Instances provide physical host isolation but carry a substantial per-hour premium, and that isolation does not inherently improve the performance or throughput of an HPC job. The AWS hypervisor already minimizes contention for modern Nitro-based instances; your bottleneck is likely vCPU, memory, or network, not noisy neighbors. Paying extra for Dedicated Instances would increase the bill without addressing the nightly, interruptible nature of the workload, making it a purely cost-negative choice.
- ✓
Use Spot Instances to take advantage of lower pricing.
Why this is correct
Spot Instances use spare EC2 capacity and are typically 60–90% cheaper than On-Demand, making them highly attractive for a nightly HPC run that can be checkpointed and resumed after an interruption. Since the batch job runs at night and is time-boxed, you can tolerate the risk of a Spot capacity reclaim by designing the workflow to restart from saved state in S3 or EFS. This directly addresses the cost problem both for compute and for the short-lived nature of the job, unlike longer-term commitments.
Quick reference
AWS S3 Storage Class Comparison
| Storage Class | Min Duration | Retrieval | Use Case |
|---|---|---|---|
| S3 Standard | None | Immediate | Frequently accessed data |
| S3 Standard-IA | 30 days | Immediate | Infrequent access, rapid retrieval |
| S3 One Zone-IA | 30 days | Immediate | Non-critical infrequent data |
| S3 Intelligent-Tiering | None | Immediate–hours | Unknown or changing access patterns |
| S3 Glacier Instant | 90 days | Milliseconds | Archive with instant retrieval |
| S3 Glacier Flexible | 90 days | Minutes–hours | Archive, flexible retrieval |
| S3 Glacier Deep Archive | 180 days | Hours | Long-term compliance archive |
Go deeper
Related to this question
About these practice questions
One of 1,169 original SOA-C02 practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
Same concept, more angles
1 more way this is tested on SOA-C02
These questions test the same concept from different angles. Work through them to make sure you can recognise it however the exam phrases it.
Variation 1. A company runs a data processing pipeline on AWS. The pipeline consists of EC2 instances that process data from an S3 bucket and write results to another S3 bucket. The processing job runs every hour and takes approximately 45 minutes. The current setup uses On-Demand instances. The SysOps administrator wants to reduce costs because the monthly EC2 bill is $5,000. The application is fault-tolerant and can handle interruptions by reprocessing data from the last checkpoint. The administrator has tested Spot Instances and found that they are interrupted about 10% of the time. The company has a strict requirement that the job must complete within 60 minutes every hour. Which solution would reduce costs while ensuring the job completes on time?
hard- A.Purchase Reserved Instances for the expected capacity to get a discount.
- B.Use a single larger On-Demand instance to complete the job faster.
- ✓ C.Use a Spot Fleet with a fallback to On-Demand if Spot capacity is not available.
- D.Use Spot Instances only and increase the number of instances to compensate for interruptions.
Why C: A Spot Fleet with a fallback to On-Demand ensures the job completes within the 60-minute window even if Spot Instances are interrupted, while reducing costs by using cheaper Spot Instances when available. Option A is wrong because Reserved Instances require a 1- or 3-year commitment and may not be cost-effective for a job that runs only 45 minutes per hour, leading to wasted capacity. Option B is wrong because using a single larger On-Demand instance would increase costs, not reduce them. Option D is wrong because using Spot Instances only, even with more instances, risks exceeding the 60-minute deadline due to the 10% interruption rate; the fallback to On-Demand is necessary to guarantee completion time.
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This SOA-C02 practice question is part of Courseiva's free Amazon Web Services certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the SOA-C02 exam.