Maximizing speedup through performance prediction for Distributed Shared Memory systems

Parallel applications executing on large system size of parallel systems achieve better speedup than on small system size. However, because of the design and implementation of the Distributed Shared Memory (DSM) system, there are some instances that large system size has no further performance improvement over small system size. It is important to determine what system size will result in the maximum speedup while all kinds of applications are running on DSM systems. In this paper rte describe rite design and implementation of we performance prediction mechanism ill our DSM system, Proteus [13], which supports node reconfiguration to adjust system size at runtime. We adopt a simple computation model and combine it with runtime information to predict the performance under different system sizes. With this work, it is possible to provide timely prediction result for the underlying system to adjust system size and thus maximize speedup.