9308900 Berman In this proposal, we outline a model and a mechanism for scheduling large-scale scientific applications on a non-dedicated heterogeneous network of computers. To obtain efficient schedules, our system will consider both domain-and application- specific information as well as system-wide characteristics. We seek a scalable and portable scheduling system with applicability to a wide variety of computational environments. We plan to use two important scientific applications with diverse computational requirements (global climate modeling and PACVHM) to test our scheduler. The initial implementation will use PVM as the underlying control mechanism. Our intent, is that the scheduler will be able to use any distributed task control as it foundation. ***
