A highly fault-tolerant quorum consensus method for managing replicated data

Abstract

The main objective of data replication is to provide high availability of data for processing transactions. Quorum consensus (QC) methods are frequently applied to managing replicated data. In this paper, we present a new QC method. The proposed QC approach has a low message overhead: 1) In the best case, each transaction operation process needs only to communicate with $O\left( {\sqrt n \log ^{1 - \tfrac{1}{{2\log _{3^2 } }}} n} \right) \left( { \approx O\left( {\sqrt n \log ^{0.208} n} \right)} \right)$ remote sites (n is the number of sites storing the manipulating data item). 2) In the worst case, each transaction operation process may be forced to communicate with $O\left( {\sqrt n \log ^{\tfrac{1}{{2\log _{3^2 } }}} n} \right) \left( { \approx O\left( {\sqrt n \log ^{0.792} n} \right)} \right)$ remote sites. Further, we can show that the proposed QC method is highly fault-tolerant. The proposed approach is also fully distributed, that is, each site in a distributed system bears equal responsibility.