SCSC 2007 START Conference Manager    

An Improved Replacement Algorithm in fault-tolerant meshes

Maryam Sadrmousavi and saina jalili

Summer Computer Simulation Conference 2007 (SCSC 2007)
San Diego, California (USA), July 15-18, 2007


Abstract

Since the failure of resources fatally affects processor allocation, a fault tolerant service is essential in the interconnection networks. In this paper, a new fault tolerant method is proposed and evaluated in the hybrid processor allocation scheme, which we have introduced in our previous work. Our task consists of two independent phases. First, the allocation process executes to allocate an efficient set of processors to the requested submesh. The second phase comes to work when the faulty nodes are detected in the allocated spaces. The selected processor allocation scheme allows jobs to be executed without waiting, provided that the number of processors is sufficient in the system and applicable to any size of requests. In addition, our fault tolerant algorithm adds redundancies (spare nodes and links) to the mesh network when the faulty nodes are detected. The replacement algorithm replaces faulty nodes with spare nodes by considering the location of the allocated submeshes in the system. Comparing results shows that the system performance, which has increased by applying the allocation scheme, can improve by using an efficient replacement algorithm.


  
START Conference Manager (V2.54.4)
Maintainer: sbranch@scs.org