There has been an increasing interest in the use of clusters of workstations connected together by high speed networks for solving large computation-intensive problems. The trend is mainly driven by the cost eectiveness of such systems as compared to large multiprocessor systems with tightly coupled processors and memories. Parallel computing on a cluster of workstations connected together by high speed networks has given rise to a range of hardware and network related issues on any given platform. Load balancing, inter-processor communication, and transport protocol for such machines are being widely studied. With the availability of cheap personal computers, workstations and networking devices, the recent trend is to connect a number of such workstations to solve computation-intensive tasks in parallel on such clusters.
The workstations can be connected using dierent network technologies such as o the shelf devices like Ethernet to specialized networks. Such networks and the associated software and protocols introduce latency and throughput limitations thereby increasing the execution time of cluster based computation. Researchers are engaged in designing algorithms and protocols to minimize the eect of this latency.
This special issue was planned to bring together the present state of the art in hardware and network requirements for running parallel applications in a network of workstations connected together by high speed networks, as well as to explore future directions of research and development. A very large number of papers were submitted indicating the public interest in the topics; after a careful review the following papers were accepted.
There is a common belief that the load should be evenly distributed among all the workstations in cluster. If a bus is used to connect the workstations in a cluster, uniform distribution of load may lead to performance degradation due to bus conΒ―icts when all the processors complete their tasks at the same time. Lin and Xie have proposed skewing of the task completion times in order to avoid bus contention. The authors have carried out analysis and simulation to show that communication conΒ―icts can be minimized in a bus based system by using a load skewing task assignment strategy.
An Ethernet switch can be used to interconnect a number of workstations in a cluster. Transport protocols normally use some form of acknowledgment to make