Which collection of design choices should a solutions architect make to achieve the maximum performance from the HPC cluster?
(Choose three.)
Ensure the HPC cluster is launched within a single Availability Zone.
Launch the EC2 instances and attach elastic network interfaces in multiples of four.
Select EC2 instance types with an Elastic Fabric Adapter (EFA) enabled.
Ensure the clusters is launched across multiple Availability Zones.
Replace Amazon EFS win multiple Amazon EBS volumes in a RAID array.
Replace Amazon EFS with Amazon FSx for Lustre.
Explanations:
Launching the HPC cluster within a single Availability Zone minimizes latency between nodes, which is essential for tightly coupled workloads.
Attaching network interfaces in multiples of four is not relevant to improving HPC cluster performance or shared file system throughput in this scenario.
Using EC2 instances with Elastic Fabric Adapter (EFA) enabled reduces latency and increases throughput, which benefits tightly coupled HPC workloads.
Spanning multiple Availability Zones would increase latency, which is detrimental to the performance of tightly coupled HPC workloads.
Using Amazon EBS volumes in a RAID array is not suitable for shared storage in HPC environments and cannot match the scalability of file storage like EFS.
Amazon FSx for Lustre is optimized for high-performance computing, providing lower latency and higher throughput than Amazon EFS for HPC applications.
I believe the answer is:
Ensure the HPC cluster is launched within a single Availability Zone.