A cluster placement group packs EC2 instances close together within a single Availability Zone, providing low latency and high network throughput (up to 10 Gbps or more) ideal for tightly coupled, high-performance computing workloads.
A cluster placement group places EC2 instances into a single Availability Zone, packing them close together on the same high-bandwidth, low-latency network fabric. This configuration enables instances to achieve non-blocking 10 Gbps network performance (up to 100 Gbps on compatible instance types) and single-digit millisecond latency. Cluster groups are ideal for tightly coupled, latency-sensitive workloads where inter-instance communication is critical, such as high-performance computing (HPC), financial simulations, and large-scale machine learning training.
All instances must be in the same Availability Zone
Use instance types that support enhanced networking and placement group awareness (e.g., C5n, M5n, R5n, P3dn)
Recommended to launch all instances in the same request or use launch templates for consistent placement
Maximum of 20 instances per placement group, though you can request a higher limit by contacting AWS Support
Use Elastic Fabric Adapter (EFA) for even lower latency and higher throughput on compatible instances
High-performance computing (HPC) clusters
Large-scale simulations (computational fluid dynamics, weather modeling)
Financial risk modeling and real-time trading analytics
Machine learning distributed training (e.g., TensorFlow, PyTorch with Horovod)
Real-time data processing with low-latency requirements
Network-bound applications requiring non-blocking 10 Gbps+ throughput