Publikation

Dal Cin, M. ; Hohl, W. ; Hönig, J. ; Pataricza, A.:
MEMSY A Modular Expandable Multiprocessor System with Fault Tolerance

Abstract:
The experimental multiprocessor system MEMSY will be described. This system was built to validate the concept of the MEMSY architecture - a scalable multiprocessor architecture based on local shared-memory. Main application areas are scientific computations with a high demand for processing power and memory capacity. In designing the hardware architecture the extensive use of standard components was a condition. The programming model of MEMSY is custom made reflecting its real hardware structure whereas th eoperating system is a Unix extension. In massively parallel systems with its complexity and large number of components the chance of a single or multiple failure is no longer negligible. It is clear that the redundancy, reconfigurability and diagnosis techniques must be incorporated at the design stage itself and not as a subsequent add-on.