Impact of Failure within a Network on Adaptive Routing Algorithm in Torus Topology

Abstract

Specific research work requires particular topology which supports proper routing algorithms with appropriate flow control schemes. Fat-Tree and Torus are the most preferred topology today. Among top 500 supercomputer top 10 computer uses these two topologies and out of top 10, 6 supercomputer uses torus topology. Special features like Scalability, maximum path diversity, increased throughput, reduced latency and enhanced performance in a multi-dimension network made torus more popular. Adaptive routing algorithm showed the promising result when used with torus topology. Even after having such a well-settled framework structure still, we are facing faults in the network. These faults can be categorized into two types depending on hardware and software called hard failures and soft failures. Researchers are able to almost eliminate soft failures like Deadlock and Livelock but the network is still getting hampered by unpredictable hard failures like link failure and node failure. In this paper, we have analyzed and compared the performance of two different adaptive routing algorithms namely Distance Vector Routing (DV) and Link-State Routing (LS) with three different types of traffic (Constant Bit Rate(CBR), Exponential & Pareto) by having the same bandwidth and packet size in normal and failure condition. A detailed simulation has been carried out using NS2. We have calculated Average Throughput, Average Delay, and Average jitter as a metric parameter for performance analysis of 3D torus topology.