SCIPVM: Parallel distributed computing on SCI workstation clusters

Abstract

Workstation and PC clusters interconnected by SCI (Scalable Coherent Interface) are very promising technologies for high performance cluster computing. Using commercial SBus to SCI interface cards and early system software and drivers, a two-workstation cluster has been constructed for initial testing and evaluation. The PVM system has been adapted to operate on this cluster using raw device access to the SCI interconnect, and preliminary communications performance tests have been carried out. Our preliminary results indicate that communications throughput in the range of 3.5 MBytes/s, and latencies 1 of 620 s can be achieved on SCI clusters. These gures are signiicantly better (by a factor of 3 to 4) than those attainable on typical Ethernet LAN's. Moreover , our experiments were conducted with rst generation SCI hardware, beta device drivers, and relatively slow workstations. We expect that in the very near future, SCI networks will be capable of delivering several tens of MBytes/s bandwidth and a few tens of microseconds latencies, and will signiicantly enhance the viability of cluster computing.