Asynchronous Consensus Without Rounds

TL;DR

This paper introduces a new class of asynchronous fault-tolerant consensus protocols that operate without rounds, reducing complexity and maintaining fault tolerance with optimal process counts.

Contribution

It provides a general specification for asynchronous consensus and presents protocols that do not rely on rounds, improving efficiency and fault tolerance.

Findings

Crash-tolerant protocols use 3f+1 processes

Byzantine-tolerant protocols use 5f+1 processes

Protocols operate without rounds, simplifying consensus mechanisms

Abstract

Fault tolerant consensus protocols usually involve ordered rounds of voting between a collection of processes. In this paper, we derive a general specification of fault tolerant asynchronous consensus protocols and present a class of consensus protocols that refine this specification without using rounds. Crash-tolerant protocols in this class use 3f+1 processes, while Byzantine-tolerant protocols use 5f+1 processes.