AWS Certified Solutions Architect – Professional (SAP-C02) — Question 266

A company has built a high performance computing (HPC) cluster in AWS for a tightly coupled workload that generates a large number of shared files stored in Amazon EFS. The cluster was performing well when the number of Amazon EC2 instances in the cluster was 100. However, when the company increased the cluster size to 1.000 EC2 instances, overall performance was well below expectations.

Which collection of design choices should a solutions architect make to achieve the maximum performance from the HPC cluster? (Choose three.)

Answer options

• A. Ensure the HPC cluster is launched within a single Availability Zone.
• B. Launch the EC2 instances and attach elastic network interfaces in multiples of four.
• C. Select EC2 instance types with an Elastic Fabric Adapter (EFA) enabled.
• D. Ensure the cluster is launched across multiple Availability Zones.
• E. Replace Amazon EFS with multiple Amazon EBS volumes in a RAID array.
• F. Replace Amazon EFS with Amazon FSx for Lustre.

Correct answer: A, C, F

Explanation

To maximize performance for a tightly coupled HPC workload scaled to 1,000 instances, launching the cluster in a single Availability Zone minimizes network latency, while utilizing Elastic Fabric Adapter (EFA) provides high-bandwidth, low-latency inter-node communication. Replacing Amazon EFS with Amazon FSx for Lustre is critical because FSx for Lustre is specifically designed for high-performance, sub-millisecond, and highly concurrent file access required by massive HPC clusters. Other options, like multiple AZs or EBS RAID arrays, either increase latency or fail to provide a distributed, concurrent shared file system for 1,000 instances.