Stretching Multi-Ring Paxos

TL;DR

This paper evaluates Multi-Ring Paxos, a scalable group communication primitive, under extreme conditions like high bandwidth, many rings, and global deployment, and proposes extensions to improve its performance.

Contribution

It provides a detailed performance analysis of Multi-Ring Paxos in challenging environments and introduces two novel extensions to enhance its efficiency.

Findings

Multi-Ring Paxos performs well under high-bandwidth and large-scale deployments.

Recovery performance is effective under peak load conditions.

Two new extensions significantly boost Multi-Ring Paxos's performance.

Abstract

Internet-scale services rely on data partitioning and replication to provide scalable performance and high availability. Moreover, to reduce user-perceived response times and tolerate disasters (i.e., the failure of a whole datacenter), services are increasingly becoming geographically distributed. Data partitioning and replication, combined with local and geographical distribution, introduce daunting challenges, including the need to carefully order requests among replicas and partitions. One way to tackle this problem is to use group communication primitives that encapsulate order requirements. This paper presents a detailed performance evaluation of Multi-Ring Paxos, a scalable group communication primitive. We focus our analysis on "extreme conditions" with deployments including high-end 10 Gbps networks, a large number of combined rings (i.e., independent Paxos instances), a large number of replicas in a ring, and a global deployment. We also report on the performance of recovery under peak load and present two novel extensions to boost Multi-Ring Paxos's performance.