The PSU PPerfDB group is developing techniques for storing, navigating, and analyzing performance data from parallel and distributed applications. We are currently focused on two goals: (1) develop a representation for resources that can accommodate performance data from a wide variety of sources and (2) develop new techniques for efficiently diagnosing parallel application performance by including historical data from previous program runs. Our testbeds are a variety of high performance clusters.

PPerfDB Publications and Presentations

• John Hoffman, Andrew Byrd, Kathryn M. Mohror, and Karen L. Karavanic, "PPerfGrid: A Grid Services-based Tool for the Exchange of Heterogeneous Parallel Performance Data," HIPS-HPGC 2005 Joint Workshop on High-Performance Grid Computing and High-Level Parallel Programming Models, in conjunction with IPDPS 2005, Denver Colorado USA, April 2005.
• Mathew E. Colgrove, Christian L. Hansen, and Karen L. Karavanic, "Managing Parallel Performance Experiments with PPerfDB," PDCN 2005, Innsbruck, Austria, February 2005.
• Kathryn Mohror and Karen L. Karavanic, "Performance Tool Support for MPI-2 on Linux," SC2004, November 2004, Pittsburgh, PA.
• Kathryn Mohror and Karen L. Karavanic, "Performance Tool Support for MPI-2 on Linux ," PSU CS Department Technical Report, April 2004.
• Kathryn Mohror and Karen L. Karavanic, "Infrastructure for Performance Tuning LAM/MPI Applications," Research Poster Exhibit, Richard Tapia Celebration of Diversity in Computing Conference, Atlanta, GA, October 2003.
• Mathew Colgrove, Christian Hansen, Brian Kearns, Suzanne Wood, and Karen L. Karavanic, "Diagnosing Parallel Application Performance with PPerfDB," Research Poster Exhibit, SC2002, Baltimore, MD, November 2002.

Masters Theses

• John Hoffman, "PPerfGrid: A Grid Services-Based Tool for the Exchange of Heterogeneous Parallel Performance Data," PSU CS Department Masters Thesis, 2004.
• Kathryn Mohror, " Infrastructure for Performance Tuning MPI Applications," PSU CS Department Masters Thesis, 2003.
• Brian Kearns, "A Performance Study of LAM and MPICH on an SMP Cluster," PSU CS Department Masters Thesis, 2002.
• Robert Jones, "Content Aware Request Distribution for High Performance Web Service: A Performance Study," PSU CS Department Masters Thesis, 2002.
• Mathew Colgrove, "Querying Geographically Dispersed, Heterogeneous Data Stores: the PPerfXchange Approach," PSU CS Department Masters Thesis, 2002.
• Christian Hansen, " Towards Comparative Profiling of Parallel Applications with PPerfDB," PSU CS Department Masters Thesis, 2001.

The PPerfMark Benchmarks

PPerfMark is a set of MPI benchmarks designed for testing performance tools for correct measured values and reasonable performance diagnoses. PPerfMark 2.2, which includes MPI-1 features, plus selected MPI-2 features, is available here for download. Download permission is granted for non-commercial uses, provided the copyright information in the code is not altered.

[Note: An earlier version of these benchmarks was called "PPerfSuite 1.0." PPerfMark 2.2 includes the complete set of PPerfSuite 1.0 tests.]

PPerfMark 2.2 tarred and gzipped