ABSTRACT
Collections of interconnected workstations are widely considered a cost-efficient alternative to supercomputers. Despite this consensus and high demand, there is no general parallel processing framework that is accepted and used by all. It is hard to adapt existing solutions because they often target a specific platform, require special networking hardware, or enforce complicated programming conventions. As a result, many organizations choose to implement custom frameworks that, for the same reasons, cannot be used by others. The effort invested in devising such frameworks compromises the cost benefit of using off-the-shelf computing resources.
