[Previous] [Top] [Next]


This thesis has presented a design process by which performance visualization developers can drastically reduce the costly and time-consuming overhead of programming, yet gain significant power and flexibility in the displays they can generate. The application of a high-level visualization methodology to general data visualization software, like IBM's Data Explorer, leads to a design technology that improves the visualization development process by increasing both the quality of the displays and the speed with which they are created. As a whole, the process lends itself to iterative design and evaluation which is required to validate a display's usability - techniques that have not been widely applied to the arena of performance visualization. Furthermore, the work being done at the University of Oregon and other research sites strongly suggests that users will demand interaction with and customization of next-generation performance displays. Existing data visualization software products are designed with these very capabilities in mind.

The work presented in this thesis can be extended in several ways. In terms of the high-level methodology developed in Chapter V, techniques that enable the specification of performance and view abstractions and their subsequent instantiations to performance and view objects are needed. This would entail automating the creation of trace and graphical transformations (i.e., the creation of data object files and visual programs) from more abstract specifications. The development of performance visualization-specific modules within Data Explorer, dedicated to certain performance visualization techniques (e.g., Kiviat diagrams and tubes, interprocessor communication displays, etc.), would make the trace and graphical transformations more independent, and consequently improve the implementation's consistency with the high-level methodology. Another area of interest is to exploit more of the graphical capabilities of the data visualization software and to see what other types of visualizations are possible. Finally, additional work is required to refine a methodology for evaluating visualizations in this environment.

The increasing sophistication and complexity of parallel computing environments will continue to present new challenges to understanding the operation and performance behavior of parallel programs. Visualization has the inherent capacity to facilitate parallel program evaluation only if its graphical data presentation capabilities, routinely used for scientific data analysis, can be effectively applied to reveal important properties of parallel program execution data. The most productive approach to developing useful visualizations for inclusion in parallel programming tools is to experiment with different visualization techniques in specific application contexts, such as those described in the three case studies documented earlier. The visualization methodology, model, and techniques proposed in this thesis are a significant step toward the integration of sophisticated graphical resources into the maturing field of parallel program and performance visualization.

Last modified: Wed Jan 20 15:14:45 PST 1999
Steven Hackstadt / hacks@cs.uoregon.edu