AWS Placement Groups
In AWS, Placement Groups are a way to influence how Amazon EC2 instances are physically placed on underlying hardware. This can improve performance, fault tolerance, or both, depending on the type of placement group you choose.
| Type | Description | Use Case |
|---|---|---|
| Cluster | Places instances close together in a single Availability Zone | Low-latency, high-throughput apps |
| Partition | Spreads instances across logical partitions on separate hardware | Large distributed systems like HDFS |
| Spread | Places instances on distinct hardware (different racks) | High availability, small instance sets |
1. Cluster Placement Group
- Instances are grouped closely in one AZ
- Provides lowest latency & highest network throughput
- Recommended for HPC (High-Performance Computing)
💡 Limitation: All instances must be in the same instance type and in a single AZ.
2. Partition Placement Group
- Instances are divided into partitions, each on separate hardware
- Fault domains: Failure in one partition doesn’t affect others
- Ideal for big data systems (like Hadoop, Cassandra)
💡 Each partition can hold many instances but is isolated from others physically.
3. Spread Placement Group
- Each instance is placed on separate underlying hardware
- Maximizes fault tolerance
- Suitable for critical small workloads (max 7 instances per AZ)
💡 Used when you want to avoid single points of failure.
📌 Notes:
- You can’t mix placement group types.
- Use the same AZ when launching into Cluster or Spread groups.
- Not all instance types are supported in placement groups.