Medium1 markMultiple Choice
Domain 2.5: Performance ObjectivesEC2PerformanceNetworkingMachine Learning

AWS SAP-C02 · Question 30 · Domain 2.5: Performance Objectives

A data analytics company runs heavy machine learning workloads on Amazon EC2 instances. The workloads require massive network throughput and low latency between instances in the same cluster. They are currently using standard EC2 networking but are experiencing network bottlenecks. What is the BEST architectural change to optimize network performance?

Answer options:

A.

Launch the instances in a Spread Placement Group and use Elastic Network Interfaces (ENIs).

B.

Launch the instances in a Cluster Placement Group and attach an Elastic Fabric Adapter (EFA) to each instance.

C.

Use AWS Transit Gateway to route traffic between the instances.

D.

Enable Enhanced Networking with the Intel 82599 Virtual Function (VF) interface.

How to approach this question

Identify the specific network adapter designed for HPC/ML workloads.

Full Answer

B.Launch the instances in a Cluster Placement Group and attach an Elastic Fabric Adapter (EFA) to each instance.✓ Correct
Launch the instances in a Cluster Placement Group and attach an Elastic Fabric Adapter (EFA) to each instance.
For tightly coupled HPC and ML workloads, a Cluster Placement Group ensures instances are physically close together. An Elastic Fabric Adapter (EFA) allows applications to bypass the operating system kernel, providing the lowest latency and highest throughput possible on AWS.

Common mistakes

Choosing Enhanced Networking (ENA) instead of EFA for ML workloads.

Practice the full AWS Solutions Architect Professional SAP-C02 Practice Exam 4

75 questions · hints · full answers · grading

More questions from this exam