Parallel and Distributed Workflows
The terms parallel and distributed can mean different things depending on the person and context. In these guides, parallel jobs mean collections of tasks that can run at the same time independently of one another. Distributed jobs, on the other hand, are those that use applications specially written to run across multiple memory spaces (typically nodes) thereby allowing them to scale up on clusters. Distributed jobs often use MPI or another distributed programming language like Chapel.
Fuzzball supports parallel jobs through “task arrays” and distributed jobs running on either MPI or PGAS-based networks.