Efficient Fault-Tolerant Adaptive Routing under an unconstrained Set of Node and Link Failures for Many Cores System On Chip

International audiencen online fault tolerant routing algorithm for 2DMesh Networks-on-Chip is presented in this work. It combinesan adaptive routing algorithmwith neighbor fault-awareness anda new traffic-balancing metric. To be able to cope with runtimepermanent and temporary failuresthat may result in messagecorruption, message loss or deadlocks, the routing algorithm isenhanced with packet retransmission and a new messagerecovery scheme.Simulation results, for various network sizes, different trafficpatterns, under an unconstrained number of node and link faults,temporary and/or permanent,demonstrate the scalability andefficiency of the proposed algorithmto tolerate multiple failureslikely encountered in deep submicron technologies. As theexperiments have shown, the proposed algorithm maintains highreliability of more than 99.38% for a 2D mesh network of 16x16and in the presence of 384 simultaneous link faults