سال انتشار: ۱۳۸۲
محل انتشار: نهمین کنفرانس سالانه انجمن کامپیوتر ایران
تعداد صفحات: ۸
Hamid sarbazi-Azad – School of Computer Science Institute for Studies inTheroretic Physics & Maths (IPM), Tehran, Iran
Hamid Reza Zarandi – Department of Computer Engineering Sharif Universiyt of Technology Tehran, Iran
This paper presents a new fault – tolerant routing algorithm in hypercube multicomputer based on the local-information of each node. In this algorithm , each node is only aware of its neighbor’s status and makes routing decision using this information and the destination address of a message. This bounded local – information , whose are numerable, causes the algorithm to be more adaptive with respect to the similar works, e.g. probability vector. Due to the low overhead of communications needed for constructing thelocal-information , dynamic faults are highly tolerable as well as static faults. Simulation results reveal that , in the case of static faults, the latency of messages is reduced up to20 percent and in the presenceof dynamic faults, the throughput of message delivery with wormhole switching is considerably improved.