A STUDY OF ADAPTIVE ROUTING PROTOCOLS FOR INJURED HYPERCUBES

Abstract

The hypercube connected multiprocessor systems are becoming more popular because of their use in real-time applications. The performance of the multiprocessor system depends largely on the communication between the processors. Hence, the routing protocols must be capable of routing the messages in the most optimal route especially when there are faults occuring in the system. This considerably decreases the delays in the system. In literature, various protocols have been proposed for faulty hypercubes. ` In this dissertation, two adoptive routing protocols, namely A l and Exhaustive Backtracking (EB) protocols have been implemented and their ability to route the message through an optimal path in the presence of faults has been studied. Also latency, header length and fault tolerating ability of these two protocols also has been studied for different sized hypercubes, for comparison. The protocol A I is found to tolerate n-1 faults in an n-cube whereas EB could tolerate any number of faults.