Resources

Defang allows you to customize the resources allocated to your services both at build-time and at run-time.

Run-time Resources

You can configure the resources available to your Defang services as required. You can configure the CPU, and memory allocated to your services as well as the number of replicas and whether or not your services requires access to GPUs.

Examples

Docker Compose

services:
  gpu-service:
    deploy:
      replicas: 3
      resources:
        reservations:
          cpus: "1.0"
          memory: 2048M
          devices:
            - capabilities: ["gpu"]

Pulumi

const service = new defang.DefangService("gpu-service", {
  deploy: {
    replicas: 3,
    resources: {
      reservations: {
        cpu: 1.0,
        memory: 2048,
        devices: [{ capabilities: ["gpu"] }],
      },
    },
  },
});

GPUs

If you require access to GPUs, you can specify this in the deploy.resources.reservations.devices[0].capabilities section of your service as in the examples above. You can learn more about this in the Docker-Compose documentation. This is the only supported value in the deploy.resources.reservations.devices section.

Build-time Resources

You can configure the memory requirements and disk space requirements for your image builds by using the shm_size property of your service's build specification.

For example,

services:
  my_service:
    build:
      context: .
      dockerfile: Dockerfile
      shm_size: 2G

info

Defang uses shm_size to configure both the memory and disk space available to your build process.

The default shm_size values for each platform are as follows. More or less may be specified.

Provider	shm_size Minimum
AWS	16G
Digital Ocean	8G
GCP	16G