Skip to main navigation Skip to search Skip to main content

When Paxos meets erasure code: Reduce network and storage cost in state machine replication

  • Shuai Mu
  • , Kang Chen
  • , Yongwei Wu
  • , Weimin Zheng
  • Tsinghua University

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

27 Scopus citations

Abstract

Paxos-based state machine replication is a key technique to build highly reliable and available distributed services, such as lock servers, databases and other data storage systems. Paxos can tolerate any minority number of node crashes in an asynchronous network environment. Traditionally, Paxos is used to perform a full copy replication across all participants. However, full copy is expensive both in term of network and storage cost, especially in wide area with commodity hard drives. In this paper, we discussed the non-triviality and feasibility of combining erasure code into Paxos protocol, and presented an improved protocol named RS-Paxos (Reed Solomon Paxos). To the best of our knowledge, we are the first to propose such a combination. Compared to Paxos, RS-Paxos requires a limitation on the number of possible failures. If the number of tolerated failures decreases by 1, RS-Paxos can save over 50% of network transmission and disk I/O. To demonstrate the benefits of our protocol, we designed and built a key-value store based on RS-Paxos, and evaluated it on EC2 with various settings. Experiment results show that RS-Paxos achieves at most 2.5x improvement on write throughput and as much as 30% reduction on latency, in common configurations.

Original languageEnglish
Title of host publicationHPDC 2014 - Proceedings of the 23rd International Symposium on High-Performance Parallel and Distributed Computing
PublisherAssociation for Computing Machinery
Pages61-72
Number of pages12
ISBN (Print)9781450327480
DOIs
StatePublished - 2014
Event23rd ACM Symposium on High-Performance Parallel and Distributed Computing, HPDC 2014 - Vancouver, BC, Canada
Duration: Jun 23 2014Jun 27 2014

Publication series

NameHPDC 2014 - Proceedings of the 23rd International Symposium on High-Performance Parallel and Distributed Computing

Conference

Conference23rd ACM Symposium on High-Performance Parallel and Distributed Computing, HPDC 2014
Country/TerritoryCanada
CityVancouver, BC
Period06/23/1406/27/14

Keywords

  • Asynchronous message passing model
  • Consensus
  • Erasure code
  • Paxos
  • State machine replication

Fingerprint

Dive into the research topics of 'When Paxos meets erasure code: Reduce network and storage cost in state machine replication'. Together they form a unique fingerprint.

Cite this