PDL Abstract

A Proof of Correctness for Egalitarian Paxos

Carnegie Mellon University Parallel Data Lab Technical Report CMU-PDL-13-111. August, 2013.

Iulian Moraru1, David G. Andersen1, Michael Kaminsky2

1Carnegie Mellon University
2 Intel Labs

This paper presents a proof of correctness for Egalitarian Paxos (EPaxos), a new distributed consensus algorithm based on Paxos. EPaxos achieves three goals: (1) availability without interruption as long as a simple majority of replicas are reachable—its availability is not interrupted when replicas crash or fail to respond; (2) uniform load balancing across all replicas—no replicas experience higher load because they have special roles; and (3) optimal commit latency in the wide-area when tolerating one and two failures, under realistic conditions. Egalitarian Paxos is to our knowledge the first distributed consensus protocol to achieve all of these goals efficiently: requiring only a simple majority of replicas to be non-faulty, using a number of messages linear in the number of replicas to choose a command, and committing commands after just one communication round (one round trip) in the common case or after at most two rounds in any case.

KEYWORDS: Distributed Consensus, Paxos

FULL TR: pdf