Fault Tolerance in Distributed Systems using Dynamic Vote Management

Abstract

There are several fault tolerant protocols for managing replicated files in the event of network partitioning due to site or communication link failures. Previously there
has been no software simulation of the voting protocols apart from just stochastic modeling. In this paper, we simulate and analyze the throughput of message transfer during the communication. We use various network topologies to compare the parameters such as throughput, no of packets received and sent during voting process .We have analyzed the effects of various packet properties. The analysis provides evidence for the conjecture that the grouping scheme is the optimal algorithm in the context of the voting protocols. We also compare the proposed genetic approach for voting assignment with random algorithm proposed by Akhil Kumar. This comparison shows that genetic voting assignment gives better availability than random algorithm.