Adaptive load distribution is necessary for parallel applications to co-exist e ectively with other jobs in a network of shared, heterogeneous workstations. We present three methods that provide such support for PVM applications. Two of these methods, MPVM and UPVM, adapt to changes in the workstation environment by transparently migrating the virtual processors (VPs) of the parallel application. A VP in MPVM is a Unix process, while UPVM de nes lightweight, process-like VPs. The third method, ADM, is a programming methodology for writing programs that perform adaptive load distribution through data movement. These methods are discussed and compared in terms of e ectiveness, usability, and performance.
Jeremy Casas, Ravi B. Konuru, Steve W. Otto, Rober