Performance Tuning

​

Display Helm Chart Values

helm show values dify/dify

​

1. Improve Resource Allocation

• You can improve the performance of Dify by adjusting the resource allocation for api, worker, and plugin_daemon services in the Helm chart values file.
• You can increase these values based on your environment and available resources.

api:
  replicas: 3
  serverWorkerAmount: 1
  resources:
    limits:
      cpu: 3000m
      memory: 10240Mi
    requests:
      cpu: 1500m
      memory: 5120Mi
worker:
  replicas: 3
  celeryWorkerAmount: 1
  resources:
    limits:
      cpu: 2000m
      memory: 10240Mi
    requests:
      cpu: 1000m
      memory: 5120Mi
plugin_daemon:
  replicas: 3
  resources:
    limits:
      cpu: 1000m
      memory: 3072Mi
    requests:
      cpu: 500m
      memory: 1536Mi

​

2. Improve External postgres Performance

• If you are using an external Postgres database, you can improve the performance by adjusting the following parameters in the postgresql.conf file:
  • max_connections: Set to 2000 or higher based on your environment.

​

3. api and worker Services

• You can set the number of workers based on your environment and available resources.

api:
  serverWorkerAmount: 1
worker:
  celeryWorkerAmount: 1

• celeryWorkerAmount: The number of Celery workers for worker service.

  • How to calculate:
    • If you are in production environment, you can set it to number of cpu cores.
    • If you are in testing environment, you can set it to 1 to save resources.
  • Reference: https://docs.celeryq.dev/en/stable/reference/cli.html#cmdoption-celery-worker-c

• serverWorkerAmount: The number of Gunicorn workers for api service.

  We recommend setting the number of workers to 1. If you set it to a value greater than 1, you need to ensure that your api service can handle the increased load.

  • How to calculate:
    • You can set it to number of cpu cores x 2 + 1.
  • Reference: https://docs.gunicorn.org/en/stable/design.html#how-many-workers