A distributed 2-fault tolerant control algorithm for reconfiguring fault-tolerant mesh arrays

A distributed control algorithm for reconfiguring fault-tolerant mesh arrays is described. This algorithm uses redundant interconnections and spare processing elements to allow two-dimensional mesh arrays to tolerate all double faults and most combinations of more than two faults. When tests show that one or more processing elements are not functioning, the algorithm isolates the faulty processing elements. Each processing element uses a controller to determine the interconnection pattern needed to isolate the faulty processing elements. Reconfiguration can, therefore, take place without the need for global fault knowledge or the use of a global controller. For arrays that use large processing elements, the hardware overhead is a very small portion of the array.< >