Skip to content

Support multiple topologies #369

@gal-revach

Description

@gal-revach

What you would like to be added?

Grove currently supports a single topology per cluster.
This request is to enable support for multiple topologies within the same cluster, and allow the user (via Grove API) to specify which topology to use. If no topology is requested - use the default grove topology.

Why is this needed?

Support for heterogeneous clusters with different topologies within the same cluster, e.g. GB200 and Vera Rubin.

Specifically for the flow of submitting Dynamo over Grove workload from Run:ai, it will help as Run:ai has the concept of node pools, where each can have a different topology attached. Once the user submits to a specific node pool, Run:ai should request this node pool's topology.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions