Fault tolerance in multiple bus networks with unbalanced resource utilization

A multiple bus network (MBN) connects a set of processors through a set of buses. In this paper, we present a method for converting any given MBN into one that is resilient to processor and bus faults. The proposed method works well in situations where some buses (or processors) are used less frequently than others. It allows the network designer to select a designated set of less frequently used buses (or processors) that can be treated as faulty, regardless of which buses (or processors) are actually faulty. This method also allows flexibility in balancing MBN's cost (degree and loading) with its performance (time to run an algorithm in a fault-prone environment). As MBNs are convenient abstractions of many systems with shared resources, this work may find applications in other settings as well.