A model for moldable supercomputer jobs

Abstract
The performance of supercomputer schedulers is influenced by the workloads that serve as their input.Realistic workloads are therefore critical to evaluatehow supercomputer schedulers perform in practice.There has been much written in the literature aboutrigid parallel jobs, i.e. jobs that require partitions of afixed size to run. However the majority of the paralleljobs in production today are moldable, i.e. jobs that canexecute on a variety of partition sizes. In this paper, wedescribe a workload model for moldable jobs, which isbased on a user survey and good analytical models.Our model can serve as the basis for the development Of performance-efficient strategies for selection of the Job partition size, as well as the basis for enhancing supercomputer schedulers to directly accept moldable request.

This publication has 10 references indexed in Scilit: