mpi4cloud

mpi4cloud is a service for running MPI-workloads on Azure, AWS, and GCP using low-cost unreliable VMs (SPOT/preemptible). mpi4cloud makes it simple to monitor the resource usage and cost of your jobs.

mpi4cloud does not provide a fully managed service, so everything happens on your cloud provider account.

The components that mpi4cloud provides are:

• a CLI for interacting with clusters
• a web application for tracking/predicting your spending, monitoring resource usage / logs, and controlling clusters
• a Python library with utilities for managing unreliable workloads