论文信息 - Analysis of the MPI-IO Optimization Levels with the PIOViz Jumpshot Enhancement

Analysis of the MPI-IO Optimization Levels with the PIOViz Jumpshot Enhancement

With MPI-IO we see various alternatives for programming file I/O. The overall program performance depends on many different factors. A new trace analysis environment provides deeper insight into the client/server behavior and visualizes events of both process types. We investigate the influence of making independent vs. collective calls together with access to contiguous and non-contiguous data regions in our MPI-IO program. Combined client and server traces exhibit reasons for observed I/O performance.

Thomas Ludwig | Stephan Krempel | Julian M. Kunkel | Michael Kuhn | Christian Lohse

[1] Thomas Ludwig,et al. Tracing the MPI-IO Calls' Disk Accesses , 2006, PVM/MPI.

[2] Allen D. Malony,et al. The Tau Parallel Performance System , 2006, Int. J. High Perform. Comput. Appl..

[3] Ludwig Thomas. RESEARCH TRENDS IN HIGH PERFORMANCE PARALLEL INPUT/OUTPUT FOR CLUSTER ENVIRONMENTS , 2004 .

[4] Robert B. Ross,et al. Using MPI-2: Advanced Features of the Message Passing Interface , 2003, CLUSTER.

[5] Oscar Naim D'Paola. Performance visualization of parallel programs , 1995 .

[6] Aric D. Blumer,et al. The Parallel Virtual File System , 1994 .

[7] Rajeev Thakur,et al. Users guide for ROMIO: A high-performance, portable MPI-IO implementation , 1997 .

[8] Barton P. Miller,et al. The Paradyn Parallel Performance Measurement Tool , 1995, Computer.

[9] William Gropp,et al. Reproducible Measurements of MPI Performance Characteristics , 1999, PVM/MPI.

[10] Thomas Ludwig,et al. Performance Evaluation of the PVFS2 Architecture , 2007, 15th EUROMICRO International Conference on Parallel, Distributed and Network-Based Processing (PDP'07).