IBM System Blue Gene Solution: Performance Analysis Tools

An IBM Redpaper publication

Note: This is publication is now archived. For reference only.

thumbnail 

Published on November 24, 2008, updated June 04, 2009

  1. .PDF (2.5 MB)

Share this page:   

IBM Form #: REDP-4256-01


Authors: Gary Lakner, I-Hsin Chung, Dr. Guojing Cong, Scott Fadden, Nicholas Goracke, David Klepacki, Jeffrey Lien, Dr. Christoph Pospiech, Seetharami R. Seelam and Hui-Fang Wen

    menu icon

    Abstract

    This IBM® Redpaper publication is one in a series of IBM documents written specifically for the IBM System Blue Gene/P Solution. The Blue Gene/P system is the second generation of a massively parallel supercomputer from IBM in the IBM System Blue Gene Solution series. The first section of this paper provides an overview of the Performance Application Programing Interface (PAPI). In this section, we describe the Blue Gene® implementation of PAPI and usage. The second part of this paper gives a detailed description of the IBM High Performance Computing Toolkit for the Blue Gene/P™ system. Included in this section is a discussion of the MPI Profiler and Tracer™ , CPU profiling, the Hardware Performance Monitor, and the Modular I/O library. The final section of this paper gives tips for tuning I/O performance and the IBM General Parallel File System™ (GPFS™ ).

    Table of Contents

    Part 1. Performance Application Programming Interface

    Chapter 1. Implementing the Performance Application Programming Interface on Blue Gene/P

    Chapter 2. PAPI usage

    Chapter 3. Automatically Available Performance Counters

    Part 2. High Performance Computing Toolkit

    Chapter 4. MPI Profiler and Tracer

    Chapter 5. CPU profiling using Xprofiler

    Chapter 6. Hardware Performance Monitoring

    Chapter 7. High Performance Computing Toolkit GUI

    Chapter 8. I/O performance

    Part 3. GPFS and I/O node

    Chapter 9. IBM General Parallel File System and I/O node performance tuning

     

    Others who read this also read