ADAPTIVE PARALLELISM AND PIRANHA

Desktop computers are idle much of the time. Ongoing trends make aggre gate LAN ''waste''-idle compute cycles-an increasingly attractive targ et for recycling. Piranha, a software implementation of adaptive paral lelism, allows these waste cycles to be recaptured by putting them to work running parallel applications. Most parallel processing is static : Programs execute on a fixed set of processors throughout a computati on. Adaptive parallelism allows for dynamic processor sets, which mean s that the number of processors working on a computation may vary, dep ending on availability. With adaptive parallelism, instead of parcelin g out jobs to idle workstations, a single job is distributed over many workstations. Adaptive parallelism is potentially valuable on dedicat ed multiprocessors as well, particularly on massively parallel process ors. One key Piranha advantage is that task descriptors, not processes , are the basic movable, remappable computation unit. The task descrip tor approach supports strong heterogeneity. A process image representi ng a task in mid-computation can't be moved to a machine of a differen t type, but a task descriptor can be. Thus, a task begun on a Sun comp uter can be completed by an IBM machine. The authors show that adaptiv e parallelism has the potential to integrate heterogeneous platforms s eamlessly into a unified computing resource and to permit more efficie nt sharing of traditional parallel processors than is possible with cu rrent systems.