An HPC workload runs on multiple Amazon EC2 instances and requires the lowest possible latency and the highest network throughput between the instances. How should a SysOps administrator place the EC2 instances to meet these networking requirements?

A cluster placement group places instances close together on the same underlying hardware, providing the lowest latency and highest network throughput, which is crucial for HPC workloads. Placing them within a single Availability Zone ensures this proximity. Spanning multiple Availability Zones would introduce network latency between instances due to physical separation. Partition placement groups are designed to spread instances across different racks to reduce the impact of hardware failures, not to optimize for low latency or high throughput between instances.