INTEGRATED VISUALIZATION OF PARALLEL PROGRAM PERFORMANCE DATA

Performance tuning a parallel application involves integrating perform ance data from many components of the system, including the message pa ssing library, performance monitoring tool, resource manager, operatin g system, and the application itself. The current practice of visualiz ing these data streams using a separate, customized tool for each sour ce is inconvenient from a usability perspective, and there is no easy way to visualize the data in an integrated fashion. We demonstrate a s olution to this problem using Devise, a generic visualization tool whi ch is designed to allow an arbitrary number of different but related d ata streams to be integrated and explored visually in a flexible manne r. We display data emanating from a variety of sources side by side in three case studies. First we interface the Paradyn parallel performan ce tool and Devise, using two simple data export modules and Paradyn's simple visualization interface. We show several Devise/Paradyn visual izations which are useful for performance tuning parallel codes, and w hich incorporate data from Unix utilities and application output. Next we describe the visualization of trace data from a parallel applicati on running in a Condor cluster of workstations. Finally we demonstrate the utility of Devise visualizations in a study of Condor cluster act ivity.