Fault-tolerant Routing Scheme for Nocs

A NoC architecture offers high reliability since it has multiple routes from the host to clients. A fault tolerant NoC framework is proposed which achieves maximum performance under fault. NoCs under fault become totally unfunctional. Hence the faulty components are handled separately which ensure that the network is not partitioned and results in high connectivity even under high fault. NoCs are a very complex network and are susceptible to a lot of faults. Faults can be temporary and permanent. Temporary faults occur due to the on-chip cross-talks and coupling noise. Permanent faults mostly occur due to worn-out devices, manufacturing defects and accelerated aging effects. These faults can be handled using router level dynamic reconfiguration schemes such as Dynamic Buffer swapping and Dynamic Mux Swapping. These methods deal with the crossbar and buffer faults which are the main sources of fault in the buffer level. The aim of these methods is to fully utilize the healthy resources in a router to mitigate faults. In the proposed DBS method, there is a threat of deadlocks. This can be handled by using the proposed recovery scheme. These methods result in high packet acceptancy rate and lower latency when compared to other routing schemes.