Computer Networks ( IF 5.6 ) Pub Date : 2021-08-28 , DOI: 10.1016/j.comnet.2021.108431 Rachel Arnold 1 , Dave Longley 2
In 1985, Fischer, Lynch, and Patterson presented the FLP Impossibility Theorem which states that it is impossible for an asynchronous system to reach consensus if at least one node fails; asynchrony prevents distinguishing between process crashes and delays. Traditionally, asynchronous consensus algorithms implement protocol adaptations to handle delays and prevent indefinite runs (e.g. coordination protocols in the form of ordered rounds). In this paper, we present a deterministic Byzantine fault tolerant asynchronous consensus algorithm called Continuity. Within this system, processes do not begin by supporting a possible decision value. Instead, Continuity utilizes logical monotonicity to build an initial configuration that is necessarily univalent, thus eliminating the assumed initial conditions of the FLP Impossibility Theorem. As such, Continuity achieves consensus in a wait-free manner.
中文翻译:
连续性:确定性拜占庭容错异步共识算法
1985 年,Fischer、Lynch 和 Patterson 提出了 FLP 不可能定理,该定理指出,如果至少一个节点发生故障,异步系统就不可能达成共识;异步防止区分进程崩溃和延迟。传统上,异步共识算法实现协议适配以处理延迟并防止无限期运行(例如,有序轮次形式的协调协议)。在本文中,我们提出了一种称为Continuity的确定性拜占庭容错异步共识算法. 在这个系统中,流程不是从支持可能的决策值开始的。相反,连续性利用逻辑单调性来构建必然是单价的初始配置,从而消除 FLP 不可能定理的假定初始条件。因此,Continuity 以无等待的方式达成共识。