Ancestral Dynamic Voting Algorithm for Mutual Exclusion in Partitioned Distributed Systems

Abstract

Data replication is a known redundancy used in fault-tolerant distributed system. However, it has the problem of mutual exclusion of replicated data. Mutual exclusion becomes difficult when a distributed system is partitioned into two or more isolated groups of sites. In this study, a new dynamic algorithm is presented as a solution for mutual exclusion in partitioned distributed systems. The correctness of the algorithm is proven, and simulation is utilized for availability analysis. Simulations show that the new algorithm, ancestral dynamic voting algorithm, improves the availability and lifetime of service in faulty environments, regardless of the number of sites and topology of the system. This algorithm also prolongs the lifetime of service to mutual exclusion for full and partial topologies especially for the situations where there is no majority. Furthermore, it needs less number of messages transmitted. Finally, it is simple and easy to implement.

A NovelN-Input Voting Algorithm forX-by-Wire Fault-Tolerant Systems

“…Because the minimum number of agreed but incorrect agreements in a given voter is desirable, the safety criterion can be defined as S = 1 − N ic / N t . Thus, S ∈ [0-1] and ideally S = 1 [ 14 , 37 , 38 ]. N d is number of disagreed results among N outputs (or the number of benign outputs).…”

“…The element performing this operation is called a voter [ 1 ]. Voting is an important operation in multichannel computation paradigm and realization of ultrareliable and real-time control systems that arbitrates among the results of N redundant variants which are used in various applications such as safety critical computer control systems (e.g., flight control systems, nuclear power station, and military applications), high reliable applications (e.g., file server applications and call processing applications) [ 1 ], highly available systems (e.g., distributed database and ad hoc networks) [ 2 ], and distributed systems (e.g., clock synchronization and Byzantine agreement). In some of these applications, in order to decrease the complexities and simplify the operations, the number of redundant modules is regarded as being small, mostly 3 or 5 modules; however, in some situations voting with fairly large number of inputs is required [ 3 , 4 ].…”

Introduction and Analysis of Optimal Routing Algorithm in Benes Networks

“….indeed in these networks the connection between the source and destination nodes is formed of multiple switching layers 1,15 . the way the layers connect to each other provide different kinds of networks.…”