Author_Institution :
Dept. of Comput. Sci., Aizu Univ., Fukushima, Japan
Abstract :
In this paper, we introduce a general fault tolerant routing problem, cluster fault tolerant routing, which is a natural extension of the well studied node fault tolerant routing problem. A cluster is a connected subgraph of a graph G, and a cluster is faulty if all nodes in it are faulty. In cluster fault tolerant routing (abbreviated as CFT routing), we are interested in the number of faulty clusters and the size of the clusters that an interconnection network can tolerate for certain routing problems. As a case study, we investigate the following k-pairwise CFT routing in n-dimensional hypercubes Hn: Given a set of faulty clusters and k distinct nonfaulty node pairs (s1 , t1), ..., (sk, tk) in Hn , find k fault-free node-disjoint paths si→ti , 1⩽i⩽k. We show that Hn can tolerate n-2 faulty clusters of diameter one, plus one faulty node for the k-pairwise CFT routing with k=1. For n⩽4 and 2⩽k⩽[n/2], we prove that H n can tolerate n-2k+1 faulty clusters of diameter one for the k-pairwise CFT routing. We also give an O(kn log n) time algorithm which finds the k paths for the mentioned problem. Our algorithm implies an O(n2 log n) time algorithm for the k-pairwise node-disjoint paths problem in Hn, which improves the previous result of O(n3 log n)
Keywords :
fault tolerant computing; hypercube networks; network routing; K-pairwise cluster; cluster fault tolerant routing; connected subgraph; fault tolerant routing; faulty node; hypercubes; interconnection network; node fault tolerant routing; node-disjoint paths; Clustering algorithms; Computer networks; Fault tolerance; Frequency; Hypercubes; Multiprocessor interconnection networks; Optical fibers; Routing; Telecommunication network reliability; Very large scale integration;