Osculant: a multiprocessor self-organizing task scheduler

The world of computing is rapidly changing. One model states that in the future a variety of mobile and/or relocatable computational assets will be rapidly configured and synergistically linked together to achieve a desired outcome. From a logistical viewpoint, these assets will be a collection of interconnected heterogeneous resources ranging possibly high-end platforms to simple man portable systems. This articulated environment must be more mobile, reconfigurable, and fault-tolerant than those currently found in common use. A new bottom-up resource scheduling paradigm, called Osculant, is being studied as the facilitating technology. The paper develops Osculant both from a conceptual and well as experimental standpoint. Simulated results are presented which indicates that Osculant can perform as well, and is generally better in balancing system resources than traditional task schedulers. Issues regarding network communication requirements, processor capabilities, logistics, scheduling details are developed and discussed. Because of self-organized framework provided by Osculant, it will also be shown that it has superior fault recovery capabilities in comparison to traditional top-down schedulers.