Abstract

Programmers of message-passing codes for clusters of workstations face a daunting challenge in understanding the performance bottlenecks of their applications. This is largely due to the vast amount of performance data that is collected, and the time and expertise necessary to use traditional parallel performance tools to analyze that data. This paper reports on our recent efforts developing a performance tool for MPI applications on Linux clusters. Our target MPI implementations were LAM/MPI and MPICH2, both of which support portions of the MPI-2 Standard. We started with an existing performance tool and added support for non-shared file systems, MPI-2 one-sided communications, dynamic process creation, and MPI Object naming. We present results usingthe enhanced version of the tool to examine the performance of several applications. We describe a new performance tool benchmark suite we have developed, PPerfMark, and present results for the benchmark using the enhanced tool.

Citations

893	Using MPI: Portable Parallel Programming with the Message-Passing Interface - Gropp, Lusk, et al. - 1999
137	Scalable performance analysis: the Pablo performance analysis environment - Reed, Roth - 1994
130	Autopilot: adaptive control of distributed applications - Ribler, Vetter, et al. - 1998
100	Dynamic program instrumentation for scalable performace tools - Hollingsworth, Miller, et al. - 1994
85	Using MPI-2: Advanced Features of the Message-Passing Interface - Gropp, Lusk, et al. - 1999
57	Toward scalable performance visualization with Jumpshot - Zaki, Lusk, et al. - 1999
37	gprof: a Call Graph Execution - Susan, Graham, et al.
37	TAU: A portable parallel program analysis environment for pC - Mohr, Brown, et al. - 1994
36	Sigma: a simulator infrastructure to guide memory analysis - DeRose, Ekanadham, et al. - 2002
34	Statistical Scalability Analysis of Communication Operations in Distributed Applications - Vetter, McCracken
29	Prophesy: an infrastructure for performance analysis and modeling of parallel and grid applications - Taylor, Wu, et al. - 2003
24	A standard interface for debugger access to message queue information in MPI - Cownie, Gropp - 1999
20	J.K.: The dynamic probe class library – an infrastructure for developing instrumentation for performance tools - Rose, Jr, et al. - 2001
17	MPI- The Complete Reference: Volume 2, The MPI Extensions - Gropp, Huss-Lederman, et al. - 1998
15	PARAVER: A Tool to Visualize and Analyze Parallel Code - Pillet, Labarta, et al. - 1995
14	Automatic Performance Analysis of MPI Applications Based on Event Traces, volume 1900 - Wolf, Mohr - 2000
11	Grindstone: A Test Suite for Parallel Performance Tools - Hollingsworth, Steele - 1996
4	MPI: The Complete Reference, Volume 1, The MPI Core, 2nd edition - Snir - 1998
4	500 Supercomputer Sites. Available at: http://www.top500.org - Top - 2003
2	An Open Cluster Environment for - “LAM - 1994
2	Science and Engineering Infrastructure for the 21st Century: The Role - Board - 2003
1	Presta Stress Test Benchmark v 1.2. Available at: http://www.llnl.gov/asci/purple/benchmarks/limited/presta - Purple
1	Prahalad “A Model for Performance Analysis - Chakravarthi, Kumar, et al.
1	Dynaprof 0.8 User’s Guide.” Available at: http://www.cs.utk.edu/~mucci/dynaprof/dynaprof.html - Mucci - 2002
1	Modeling the Earth’s Atmosphere.” NASA Goddard Space Flight Center. Presented at SC2002. Information available at: http://www.nas.nasa.gov/About/Media/SC02 /Goddard/goddard.html - Putman, Chern, et al.
1	Distributed Terascale Facility to Commence with $53 Million NSF Award.” TeraGrid News Release. Available at: http://www.teragrid.org/news/news01/080901_nsf.html - Project
1	Dynaprof 0.8 User’s Guide.” Available at - Mucci - 1996