Routing in modular fault tolerant multiprocessor systems

The authors consider a class of modular multiprocessor architectures in which spares are added to each module to cover for faulty nodes within that module, thus forming a fault tolerant basic block (FTBB). The goal is to preserve the logical adjacency between active nodes by means of a routing algorithm which delivers messages successfully to their destinations. Two phase routing strategies are introduced that route messages first to their destination FTBB, and then to the destination nodes within the destination FTBB. This strategy may be applied to a variety of architectures including binary hypercubes and 3-D tori. In the presence of f faults in these systems. It is shown that the worst case length of the message route is max( sigma +f, (K+1) sigma )+M, where sigma is the shortest path in the absence of faults, and M and K are the numbers of primary nodes and spare nodes in a FTBB, respectively. The average routing overhead is much lower than the worst case overhead.<<ETX>>