state machine operator

state machine orchestrator for the Mummi workflow

An HPC ensemble is an orchestration of jobs that can ideally be controlled by an algorithm. Most resource managers expect workflows that have resource needs hard-coded, and a lot of manual orchestration. This effort aims to design a workflow tool that is more akin to a state machine, and responds to different events. Instead of hard coding specific applications, we allow for them to be defined dynamically. Some assumptions we make:

• An order of steps, A->B, understands how to handle output from the previous step. E.g., if we package up the output of A and give it to B in a known working directory, B knows what to do.
• Similarly, B knows that whatever is placed in that working directory will be provided to the next step.

This project will be intended to run in Kubernetes, because we are developing user-space Kubernetes for our HPC clusters and I want to start simple.

License

HPCIC DevTools is distributed under the terms of the MIT license. All new contributions must be made under this license.

See LICENSE, COPYRIGHT, and NOTICE for details.

SPDX-License-Identifier: (MIT)

LLNL-CODE- 842614