Process migration subsystem for a workstation-based distributed systems

Workstation-based distributed computing environments are getting popular in both academic and commercial communities due to the continuing trend of decreasing cost/performance ratio and rapid development of networking technology. However, the work load on these workstations is usually much lower than their computing capacity, especially with the ever-increasing computing power of new hardware. As a result, the resources of such workstations are often under-utilized and many of them are frequently idle. A preemptive process migration facility can be provided, in such a distributed system, to dynamically relocate running processes among the component machines. Such relocation can help cope with dynamic fluctuations in loads and service needs, improve the system's fault tolerance, meet real-time scheduling deadlines, or bring a process to a special device. This paper presents a process migration subsystem for tolerating process and node failures on a workstation based environment. The design and implementation of the subsystem are also discussed.