Analyzing the Workload of Scientific Visualization Tools: A Case Study

In this paper we present a workload characterization analysis of TIPSY, which is a typical interactive package for scientific visualization and analysis. Tools like this are gaining increasing importance as parallel scientific computations produce increasing amounts of data, raising the need for efficient visualization and data mining. Our experiments show that, although TIPSY has high processing requirements, its main bottleneck is paging. This is attributed to main memory limitations and to contention by other processes. The analysis of processing and I/O behavior yields patterns which differ from those derived by studies of parallel scientific applications. Extrapolating our results for cases involving larger input file sizes reveals that paging will remain a bottleneck, even for workstations of higher performance than the ones used here. Therefore, performance enhancements based on parallelization and remote memory paging should be examined. Key Words: Interactive scientific visua..