KEP-2170: Create MPI Runtime

[andreyvelich]

Related: #2170

As part of this KEP, we will migrate to the MPI V2 implementation.

We should add support for the MPI Runtime.

/area runtime

[tenzen-y]

Note that we need to extend the KEP-2170 for the MPI before we implement anything.

[tenzen-y]

Note that we need to extend the KEP-2170 for the MPI before we implement anything.

Oh, we already added the design for the MPI here: https://github.com/kubeflow/training-operator/tree/master/docs/proposals/2170-kubeflow-training-v2#the-mpi-spec-api

NVM

[andreyvelich]

Note that we need to extend the KEP-2170 for the MPI before we implement anything.

Oh, we already added the design for the MPI here: https://github.com/kubeflow/training-operator/tree/master/docs/proposals/2170-kubeflow-training-v2#the-mpi-spec-api

NVM

Once we will be ready to implement MPI runtime, we should probably update this ClusterTrainingRuntime: https://github.com/kubeflow/training-operator/tree/master/docs/proposals/2170-kubeflow-training-v2#mpi-runtime.

It might have incorrect values, since we didn't get a chance to finalize it.

[tenzen-y]

Note that we need to extend the KEP-2170 for the MPI before we implement anything.

Oh, we already added the design for the MPI here: https://github.com/kubeflow/training-operator/tree/master/docs/proposals/2170-kubeflow-training-v2#the-mpi-spec-api
NVM

Once we will be ready to implement MPI runtime, we should probably update this ClusterTrainingRuntime: https://github.com/kubeflow/training-operator/tree/master/docs/proposals/2170-kubeflow-training-v2#mpi-runtime.

It might have incorrect values, since we didn't get a chance to finalize it.

That sounds good to me.