ExamPark
Course menuSOA-C03

Exams / SOA-C03 / Study guide / EC2 performance and right-sizing

EC2 performance and right-sizing: SOA-C03 study notes and practice questions

What to know about EC2 performance and right-sizing for the SOA-C03 exam, from the Monitoring, Logging, Analysis, Remediation, and Performance Optimization domain (22% of the exam guide), then 1 free question to test it. Part of the SOA-C03 study guide.

Last verified against current AWS documentation ·How we write our questions

What to know for the exam

Compute performance questions ask why an instance is slow, or how to make it faster or cheaper.

Right-sizing

  • AWS Compute Optimizer recommends instance types for EC2, Auto Scaling groups, EBS, Lambda, ECS on Fargate and RDS, from CloudWatch history (14 days by default, up to 93 days with enhanced infrastructure metrics).
  • It can only consider memory if the CloudWatch agent publishes memory metrics.
  • Change an EBS-backed instance’s type by stopping it, changing the type and starting it.

Burstable instances (T family)

  • Earn CPU credits below a baseline and spend them above it.
  • When credits run out in standard mode, CPU drops to the baseline; unlimited mode (default for T3 and later) keeps bursting and charges for surplus credits.
  • A steady high CPU load belongs on a fixed-performance family (M, C), not T.
  • Watch CPUCreditBalance and CPUSurplusCreditBalance.

Choosing instances

Need Family
Balanced M, T
CPU-bound C
Memory-bound R, X
Fast local storage I, D
GPUs / accelerators P, G, Inf, Trn

Graviton (Arm) instances usually give better price performance for supported software.

Networking and placement

  • Enhanced networking (ENA) gives higher bandwidth and lower latency; Elastic Fabric Adapter adds low-latency HPC networking.
  • Cluster placement groups pack instances close together for the lowest latency between them.
  • Instance types have network bandwidth limits; a bigger instance may be the fix for a network bottleneck.

Common traps

  • Right-sizing on CPU alone. Memory needs the agent.
  • A T instance throttled after a few hours of constant load: credits ran out.
  • Changing instance type while it runs. Stop it first.

Try the 1 practice question

Practice EC2 performance and right-sizingTake the free practice exam

Question 1

Medium ·Link to this question (question 1)

A logistics broker uses AWS Compute Optimizer to right-size its EC2 fleet. For its route-planning servers, Compute Optimizer recommends a smaller instance type with half the memory, but the application owners report that memory usage on these servers is regularly above 70 percent.

The operations team wants recommendations that reflect the servers' real memory use.

What should a CloudOps engineer do?

Choose one answer.

  1. A. Install the CloudWatch agent on the servers to publish memory utilization for Compute Optimizer to analyze.
  2. B. Turn on enhanced infrastructure metrics so that Compute Optimizer analyzes up to three months of utilization history.
  3. C. Turn on detailed monitoring on each instance so that Compute Optimizer receives one-minute utilization data.
  4. D. Use the Trusted Advisor Low Utilization Amazon EC2 Instances check instead, because it includes memory in its analysis.
Show the answer and explanation

Answer: A

Explanation

Compute Optimizer only knows what CloudWatch knows. EC2 reports CPU, network and disk I/O from the hypervisor, but memory needs the CloudWatch agent inside the instance. Once memory metrics are published, Compute Optimizer uses them and its recommendations account for memory-heavy workloads.

Study

Want more on EC2 performance and right-sizing? The 5 full-length practice sets have 5 more EC2 performance and right-sizing questions, 325 in all, none of them in the free exam. One payment of $14.99, no subscription. See the practice sets

Study EC2 performance and right-sizing

More Monitoring, Logging, Analysis, Remediation, and Performance Optimization topics