This paper presents a cluster-based PMBK multiprocessor system (CPMBK) and
analyses its performance of bandwidth. This interconnection network offers
advantages such as more flexibility. Two situations are considered: (1) a p
rocessor is equally likely to address any one of the memory modules, (2) a
processor has a few favorite memory modules. The results derived show that
the cluster-based PMBK multiprocessor system performs well and is fairly cl
ose to a typical partial multiple bus system with K classes (PMBK) for high
rates of favorite requests. The number of connections of this system is al
so evaluated and compared with those of other systems. It shows that the cl
uster-based PMBK system can reduce the network cost and complexity without
significantly reducing the performance compared with those of crossbar, PPM
B, PMBK and cluster-based crossbar systems.