Node-to-node Cluster Fault Tolerant Routing in Star Graphs

 

Qian-Ping Gu, Shietung Peng

 

Information Processing Letters Vol. 56, pp. 29-35, 1995

 

Abstract

In this paper, we give an algorithm which, given at most faulty clusters of diameter at most 2, and non-faulty nodes s and t in G, finds a fault-free path of length in O(|F|+n) time and a fault-free path of length at most in O(n) time, where is the diameter of and F is the set of total faulty nodes.

 

  1. CFT routing problem

Cluster fault tolerant(CFT) routing is a natural extension of conventional node fault tolerant routing mode. In CFT routing, routing problems are considered under the situation that a certain number of cluster of G failed. For a particular routing problem in an n-connected graph, how many faulty clusters and how large of those clusters can be tolerated are studied in CFT routing.

 

Lemma 2. For distinct nodes and in the same port-set of a substar,

 

Lemma 4. For any node in and any , there are disjoint paths of length at most 3 that connect s to distinct nodes in .

 

 

Example. Let and . Then the paths are:

 

Lemma 6. Given and with and , C can block at most one of the n-2 paths , , given in Lemma 4.

 

 

 

2. Routing Algorithm

Algorithm

Input: Fault clusters , non-faulty nodes s and t in .

Output: A fault-free path

begin

For each, , find the center node of ;

For and ;

Find a substar , s.t. and ;

Enumerate the disjoint paths from to

to find the fault-free routing path of length at most 3.

Enumerate the disjoint paths from to

to find the fault-free routing path of length at most 3.

Connect and in by a fault-free shortest path;

end

 

Lemma 7. Given at most faulty nodes and non-faulty nodes and in , a fault-free path of length at most that connects and can be constructed in time.

 

Lemma 8. Given a set F of faulty clusters with |F| and , and two distinct non-faulty nodes s and t in , Algorithm I finds a fault-free path of length at most in O(|F|+n) time and finds a fault-free path of length at most in time.

 

Proof: Case(1): if , then the path has been found.

Case(2): if . Since the faulty nodes in are the nodes , where is the center node of some cluster , and is fault-free, is fault-free.

Case(3): assume that and , and all the paths are blocked

Therefore, there is no center node of any faulty cluster in the substar which contains . We have nodes and , , are fault-free. Similarly, is also fault-free.

Therefore, It takes to find the center nodes of faulty clusters. By checking the last digits of and , and can be computed in time. Thus, the time complexity of the algorithm is .

It is known that given any nodes and in , there are node disjoint paths of length at most connecting and . Therefore, a fault-free path of length at most in time.

 

Da Created by: Tung-Hsing Liu

te: Mar. 1 1997

e-mail: para@ibm15.math.nsysu.edu.tw