ADAPTIVE AND FAULT-TOLERANT MESSAGE ROUTING USING DISTRIBUTED RECOVERY BLOCK APPROACH
Gul N. Khan, Wei Gu · 1999
This paper describes an adaptive and fault-tolerant message routing technique that incorporates a distributed recovery block (DRB) scheme. The reliability requirements of inter-processor communication for parallel and distributed computer systems are increasing. The message routing algorithm, presented here, ensures the delivery of every message as long as a healthy path exists between source and destination nodes. Fault tolerance is achieved by employing distributed recovery block approach, which consists of two try-block nodes with an acceptance test. The primary try-block delivers messages, whereas the alternate is ready to take over if the primary fails. We are focusing on store-and-forward message propagation strategy, however the method can also be employed to establish a reliable path for wormhole or virtual cut through routing. The method has been investigated and implemented for 2D-mesh, torus and hypercube network topologies for varying network traffic.