Table of contents

Architecture for Watson Studio Local

Watson Studio Local utilizes three separate features for cluster management, storage management, and execution.

Private cloud components

Watson Studio Local runs on a Kubernetes cluster of servers, which are composed of the following components:

Control plane (master)
For cluster management and high availability.
  • Requires three master nodes to manage the entire cluster.
  • Uses etcd as a key value store that persists the cluster state and stores metadata about cluster service deployment and health.
  • Uses Prometheus for monitoring and Elk for logging.
For data stores and storage management.
  • Uses either NFS (required for production deployments and can also be used for test deployments) or GlusterFS (for test deployments only) for storage management.
  • Uses IBM Cloudant DB as a service meta database.
  • Uses Redis as the in-memory database
  • Uses Elasticsearch DB for logs.
Runs data science related services.
Figure 1. Architecture for a minimum of four nodes
Five node diagram
Figure 2. Architecture for a minimum of seven nodes
Eight node diagram