Fault-Tolerant Metacomputing for Defense Applications

Parallel and distributed processing on "metacomputers" formed by networking scalable parallel computers with COTS workstations and PCs is clearly the wave of the future for high-performance computing. The reasons are obvious: cost and performance. The Department of Defence can benefit greatly, provided that the real-time and fault-tolerance requirements of key defense applications can be met. This project will design and assess a complete hierarchical environment supporting real-time, fault-tolerant applications on metacomputers. At the bottom, the environment will exploit basic system infrastructure for real-time support, basic group and communication services, information about hardware and software status, and fault recovery services. Above that will be a scalable, reliable global store suitable for use by the operating system and by applications to preserve critical state information. Exploiting the store will be tools for monitoring and control and job/resource scheduling, both able to work cooperatively and intelligently to run real-time parallel applications adaptively, efficiently, and correctly, even in the presence of failures. Since other projects are developing the low-level infrastructure, the focus of this project will be on the high-level portions of the hierarchy, particularly the virtual store and monitoring tool, and on design of a suitable API for development of fault-tolerant applications.