Performance analysis for the communication of parallel applications on a scalable workstation cluster with a concurrent network architecture

摘要

Today a lot of possibilities exist to create cost-efficient parallel or distributed architectures to solve a lot of special problems with a higher performance than on a single processor machine. Nevertheless it is often very difficult to predict the performance improvements achieved by a secial constructed hardware with its on system software before its realization. For the large class of SPMD-applications it shall be shown exemplary in which way the expected performance can be calulated in an easy way.