Reduce latency by using compact placement policies

This document explains how to reduce network latency among a job's VMs by specifying a compact placement policy.

Having VM hosts that are physically closer to each other within a zone can reduce network latency between VMs. This performance benefit can be especially useful for jobs with tightly coupled tasks, such as tasks that communicate using MPI libraries. When you create a job, you can optionally require a job's VMs to be located closer to each other by specifying a compact placement policy. Otherwise, default VM placement typically distributes VMs across separate hosts to help minimize the impact of power failures.

To learn more about compact placement policies, see About placement policies in the Compute Engine documentation.

Before you begin

• If you haven't used Batch before, review Get started with Batch and enable Batch by completing the prerequisites for projects and users.

• To get the permissions that you need to create a job, ask your administrator to grant you the following IAM roles:

  • Batch Job Editor (roles/batch.jobsEditor) on the project
  • Service Account User (roles/iam.serviceAccountUser) on the job's service account, which by default is the default Compute Engine service account

  For more information about granting roles, see Manage access.

  You might also be able to get the required permissions through custom roles or other predefined roles.

Restrictions

In addition to the general restrictions for compact placement policies, Batch also has the following restrictions:

• When you specify for a job to use a compact placement policy, Batch automatically creates and deletes a compact placement policy for the job's VMs. You can't specify an existing compact placement policy for a job.
• You can't specify a number of VMs for the compact placement policy. The compact placement policy for a job applies regardless of how many VMs the job runs on.

Create and run a job that uses a placement policy

You can specify for a job to use a compact placement policy by setting the collocation field to COLLOCATED using the gcloud CLI or Batch API.

gcloud

1. Create a JSON file that specifies the job's configuration details and sets the collocation field to COLLOCATED.

   For example, to create a basic script job that specifies a compact placement policy, create a JSON file with the following contents:

   {
     "taskGroups": [
       {
         "taskSpec": {
           "runnables": [
             {
               "script": {
                 "text": "echo Hello world from task ${BATCH_TASK_INDEX}"
               }
             }
           ]
         },
         "taskCount": 3
       }
     ],
     "allocationPolicy": {
       "placement": {
         "collocation": "COLLOCATED"
       }
     },
     "logsPolicy": {
       "destination": "CLOUD_LOGGING"
     }
   }
   

2. To create and run the job, use the gcloud batch jobs submit command:

   gcloud batch jobs submit JOB_NAME \
     --location LOCATION \
     --config JSON_CONFIGURATION_FILE
   

   Replace the following:

   • JOB_NAME: the name of the job.

   • LOCATION: the location of the job.

   • JSON_CONFIGURATION_FILE: the path for a JSON file with the job's configuration details.

API

Make a POST request to the jobs.create method that sets the collocation field to COLLOCATED.

For example, to create a basic script job that specifies a compact placement policy, make the following request:

POST https://batch.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/jobs?job_id=JOB_NAME

{
  "taskGroups": [
    {
      "taskSpec": {
        "runnables": [
          {
            "script": {
              "text": "echo Hello world from task ${BATCH_TASK_INDEX}"
            }
          }
        ]
      },
      "taskCount": 3
    }
  ],
  "allocationPolicy": {
    "placement": {
      "collocation": "COLLOCATED"
    }
  },
  "logsPolicy": {
    "destination": "CLOUD_LOGGING"
  }
}

Replace the following:

• PROJECT_ID: the project ID of your project.

• LOCATION: the location of the job.

• JOB_NAME: the name of the job.

What's next

• If you have issues creating or running a job, see Troubleshooting.
• View jobs and tasks.
• Learn about more job creation options.