Compact placement policy and workload policy overview

This document provides an overview and comparison of compact placement policies and workload policies. Both policies let you configure the placement of virtual machine (VM) instances to minimize network latency. Use compact placement policies for instances that are created individually or in bulk, and use workload policies for managed instance groups (MIGs).

By default, you manage the location of your VMs only by specifying their zones. When you use future reservations or flex-start to request machine types that support Cluster Director, the VM resources that you receive are densely colocated by default. However, you might want to place specific VMs more closely together to optimize inter-VM performance in each application. To place VMs closer together, you can apply compact placement policies to VMs or high-throughput workload policies to MIGs.

Compact placement policy for VMs

When you apply compact placement policies to your VMs, Compute Engine makes best-effort attempts to create VMs as close to each other as possible. If your application is latency-sensitive and you want the VMs to be as close together as possible (maximum compactness), then specify the maxDistance field ( Preview ) when creating a compact placement policy.

For more information, see About compact placement policies in the Compute Engine documentation.

Workload policy for MIGs

A workload policy lets you specify the type of the workload that you want to run on your infrastructure. You can also specify the physical properties of the underlying infrastructure, such as the VM placement, to best match the specified type.

You make the following configuration settings for a workload policy:

• Workload type( type ): for high throughput ( high-throughput ) workloads, Compute Engine makes best-effort attempts to place VMs as close as possible to each other. The VM placement depends on the machine type and zone availability of the VMs.

• Additional requirement for using strict colocation or accelerator topology of VMs. You can specify one of the following:

  • Strict colocation of VMs( maxTopologyDistance ): to achieve granular, low-latency network performance. A strict colocation means that, in addition to the best-effort to place your VMs as close to each other as possible, you can further specify the maximum distance between the VMs. If the strict colocation requirement is not met due to capacity constraints, the MIG doesn't create the VMs.

  • Accelerator topology of VMs( acceleratorTopology ): to achieve high performance for distributed workloads which run across multiple VMs that use a specialized inter-accelerator network configuration—for example, A4X VMs that use NVLink Domains for GPUs.

Comparison of compact placement policy and workload policy

The following table summarizes the differences between compact placement policies and workload policies:

Comparison of maximum distance values

A lower maximum distance value ensures closer VM placement, but it also increases the chance that some VMs won't be created.

The following table shows the machine series and number of VMs that each maximum distance value supports:

What's next

• Create an instance
• Create instances in bulk
• Create a MIG