Network Placement Groups
You'll learn how to use placement groups to achieve ultra-low latency and high throughput between compute instances in the same region.
What Are Network Placement Groups?
A placement group is a logical grouping of instances within a single availability zone that are physically placed close together on the underlying hardware. Think of it like reserved seating at a concert—instead of your group being scattered throughout the venue, you're all seated together in the same section for better communication.
When instances are part of a placement group, the cloud provider ensures they're located on the same or nearby racks in the data center, connected by high-speed, low-latency network fabric. This physical proximity dramatically improves network performance between instances.
Why Use Placement Groups?
Standard instance deployment spreads your resources across different physical hardware for resilience. But some workloads—like high-performance computing (HPC), distributed databases, or real-time analytics—need instances to communicate with minimal network overhead.
Placement groups deliver:
- Lower latency: Sub-millisecond network delays between instances
- Higher bandwidth: Full bisection bandwidth (often 10-100 Gbps) between group members
- Consistent performance: Reduced network jitter and packet loss
Trade-offs to Consider
By concentrating instances in one physical location, you sacrifice some fault tolerance. If that hardware rack experiences issues, all instances in the group may be affected. This is the opposite of spreading instances across multiple availability zones for high availability.
Use placement groups when network performance matters more than maximum fault isolation—but always architect for failure elsewhere in your stack.
Key Takeaway: Placement groups physically co-locate instances in a single availability zone to minimize network latency and maximize throughput, ideal for tightly-coupled workloads that need rapid inter-instance communication.