Fault Tolerance in Hypercube Systems and Implementation of Fault Tolerant Algorithms Kayhan ERCİYEŞ 1 Novruz ALLAHVERDİ 2 A.Belma ŞAHİN 1 1 Ege University, International Computing Institute, İzmir, Turkey e-mail : erciyes@ube.ege.edu.tr , sahin@alpha.ube.ege.edu.tr 2 Selçuk University, Faculty of Technical Education, Dept. Of Electrical and Computer Education, Konya, Turkey e-mail : noval@alaeddin.cc.selcuk.edu.tr Abstract : In this study, we‘ve analyzed and implemented three different algorithms developed for fault-tolerant routing in hypercube systems. First algorithm[1] proposes an approach to determine the shortest path between the source and the destination nodes in a faulty (or non-faulty) hypercube. Second algorithm[2] has a different approach to hypercube node structure and proposes that direct use of sharp product is not sufficent to discard only computational part (processor and memory) when only this part of a node is faulty. The procedures proposed within this algorithm, allow to obtain a set of extended fault-free subcubes which is a beginning set for further manipulation in hypercube multiprocessors and to increase the reliability of such systems. Third algorithm[3] proposes a procedure for definition of neighborhood and fault-free subcubes in a fault-free subcubes in a faulty-hypercube. We’ve implemented these algorithms in UNIX and in Intel iPSC/860 Simulation platforms. Simulations are done for static and dynamic routing. Performances are compared for path availability percentages and time used to find path. Keywords : Hypercube, fault-tolerance, shortest-path, routing, cube algebra, computational part, communicational part, iPSC