Abstract
This paper describes a kernel scheduling algorithm that is based on co-scheduling principles and that is intended for parallel applications running on 1000 cores or more. Experimental results for a Linux implementation on a Cray XT5 machine are presented. The results indicate that Linux is a suitable operating system for this new scheduling scheme, and that this design provides a dramatic improvement in scaling performance for synchronizing collective operations at scale.
| Original language | English |
|---|---|
| Pages (from-to) | 136-145 |
| Number of pages | 10 |
| Journal | International Journal of High Performance Computing Applications |
| Volume | 26 |
| Issue number | 2 |
| DOIs | |
| State | Published - May 2012 |
Funding
The author gratefully acknowledges the computer resources and assistance provided by the National Center for Computational Sciences at Oak Ridge National Laboratory. I would also like to express my gratitude to ORNL’s Don Maxwell and David Dillow for all of their assistance in getting the Jaguar and Rizzo machine prepared for this work. Further thanks are due to the entire Colony team for helping to make this work a reality. Finally I would like to thank our funding sponsors. The submitted manuscript has been authored by a contractor of the US Government under Contract No. DE-AC05-00OR22725. Accordingly, the US Government retains a non-exclusive, royalty-free license to publish or reproduce the published form of this contribution, or allow others to do so, for US Government purposes. The work described here was performed at Oak Ridge National Laboratory. It was supported in part by the Department of Energy FastOS II program (Lab 07-23, see http://www.fastos2.org/ ). Computer allocation support was provided by the Department of Energy INCITE award and an NCCS Director’s Discretion award.
Keywords
- co-scheduling
- kernel scheduling
- operating system interference
- operating system noise