The Community for Technology Leaders
SC Conference (2003)
Phoenix, Arizona
Nov. 15, 2003 to Nov. 21, 2003
ISBN: 1-58113-695-1
pp: 10
Robert Blackmore , International Business Machines Corporation, Armonk, NY
Mark Roberts , Atomic Weapons Establishment, Aldermaston Reading, UK
Shawn Dawson , Lawrence Livermore National Laboratory, Livermore, CA
William Tuel , International Business Machines Corporation, Armonk, NY
Paul Tomlinson , Atomic Weapons Establishment, Aldermaston Reading, UK
Jeffrey Fier , International Business Machines Corporation, Armonk, NY
Larry Brenner , International Business Machines Corporation, Armonk, NY
Brian Maskell , Atomic Weapons Establishment, Aldermaston Reading, UK
Rob Neely , Lawrence Livermore National Laboratory, Livermore, CA
Terry Jones , Lawrence Livermore National Laboratory, Livermore, CA
Patrick Caffrey , International Business Machines Corporation, Armonk, NY
ABSTRACT
A parallel application benefits from scheduling policies that include a global perspective of the application's process working set. As the interactions among cooperating processes increase, mechanisms to ameliorate waiting within one or more of the processes become more important. In particular, collective operations such as barriers and reductions are extremely sensitive to even usually harmless events such as context switches among members of the process working set. For the last 18 months, we have been researching the impact of random short-lived interruptions such as timer-decrement processing and periodic daemon activity, and developing strategies to minimize their impact on large processor-count SPMD bulk-synchronous programming styles. We present a novel co-scheduling scheme for improving performance of fine-grain collective activities such as barriers and reductions, describe an implementation consisting of operating system kernel modifications and run-time system, and present a set of empirical results comparing the technique with traditional operating system scheduling. Our results indicate a speedup of over 300% on synchronizing collectives.
INDEX TERMS
CITATION
Robert Blackmore, Mark Roberts, Shawn Dawson, William Tuel, Paul Tomlinson, Jeffrey Fier, Larry Brenner, Brian Maskell, Rob Neely, Terry Jones, Patrick Caffrey, "Improving the Scalability of Parallel Jobs by adding Parallel Awareness to the Operating System", SC Conference, vol. 00, no. , pp. 10, 2003, doi:10.1109/SC.2003.10024
105 ms
(Ver )