Bringing Paxos Consensus in Multi-agent Systems
Bringing Paxos Consensus in Multi-agent Systems
复制标题
在多代理系统中引入 Paxos 共识
DOI：
--
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
C. Bǎdicǎ
中科院分区：
文献类型：
--
作者：
A. Mocanu;C. Bǎdicǎ
达成共识一直被认为是分布式系统中最重要的问题之一。能够在失败的情况下这样做是由Paxos家族的算法解决的，该算法能够保证安全性，同时概率上令人满意的进展。Paxos基于角色的方法使其成为使用自治代理实现的理想候选者，这些代理可以动态地承担其部分。本文旨在通过描述Jade平台下的系统架构和实现，将基本的Paxos容错算法引入到多智能体世界中，并给出了在不利情况下收敛的有价值的实验结果。
Reaching consensus has long been regarded as one of the most important problems in distributed systems. Being able to do so under failures is addressed by the Paxos family of algorithms which is able to guarantee safety, while probabilistically satisfying progress. The role-based approach of Paxos makes it an ideal candidate for implementation using autonomous agents which can dynamically assume their part(s). This paper aims to bring the basic Paxos fault-tolerant algorithm into the multi-agent world by describing a system architecture and implementation under the Jade platform, and presents valuable experimental results for convergence in unfavorable cases.