Synthesizing Network Requirements Using Parallel Scientific Applications

We synthesize the link bandwidth requirement for a binary hypercube topology using a set of five scientific applications. We use an execution-driven simulator called SPASM to collect data points for system sizes that are feasible to be simulated. These data points are then used in a regression analysis for projecting the link bandwidth requirements for larger systems. These requirements are projected as a function of the following system parameters: number of processors, CPU clock speed, and problem size. These results are also used to project the link bandwidths for other network topologies. A significant contribution of our study is in quantifying the link bandwidth that has to be made available to tolerate a given amount of network overhead in an application. Our results show that typical link bandwidths (200-300 MBytes/sec) found in current commercial parallel architectures (such as Intel Paragon and Cray T3D) would have fairly low network overhead for the scientific applications considered in this study. For two of the applications, this overhead is negligible. For the other applications, this overhead is about 30% of the execution time provided the problem sizes are increased commensurate with the processor clock speed