The harmonized consensus protocol in distributed systems

Abstract

The goal of the consensus protocol is to make all non-faulty processors agree on a common value despite the interference of faulty processors. In the classical consensus problem, each processor can propose an initial-value and obtain a commonly agreed consensus value through message exchange. However, if the initial-value of any processor is shared by only a small number of other processors, this initial-value will not become the consensus value. This is because initial-values once determined cannot be changed in the classical consensus protocol. In this study, we propose a new consensus protocol named harmonized consensus protocol (HCP) to solve this problem. HCP has a mechanism that allows each processor to change its initial-value based on the messages it has collected during message exchange (i.e., PrfRule rules). Hence, it can increase the chances that all processors agree on a consensus value they prefer and accept. Moreover, HCP also has the mechanism for judging the sufficiency of messages collected (i.e., MajChk rule). It terminates message exchange when processors have collected enough messages for obtaining a consensus value. This also means that HCP can reduce the number of rounds of message exchanges and further lower the time and space complexity of reaching consensus. The performance evaluation results confirm that the proposed HCP can deliver a higher preference match rate with lower time and space complexity.