通信学报 ›› 2014, Vol. 35 ›› Issue (4): 166-173.doi: 10.3969/j.issn.1000-436x.2014.04.019

• 学术通信 • 上一篇    下一篇

基于再生码的云存储系统——Ustor

柳青,冯丹,李白   

  1. 华中科技大学 计算机科学与技术学院,湖北 武汉 430074
  • 出版日期:2014-04-25 发布日期:2017-07-03
  • 基金资助:
    国家重点基础研究发展计划(“973”计划)基金资助项目;国家自然科学基金资助项目;国家自然科学基金资助项目;国家高技术研究发展计划(“863”计划)基金资助项目;国家科技支撑计划基金资助项目;中央高校基本科研业务费基金资助项目;电子信息产业发展基金资助项目

Ustor: cloud storage system based on regenerating codes

Qing LIU,Dan FENG,Bai LI   

  1. School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074, China
  • Online:2014-04-25 Published:2017-07-03
  • Supported by:
    The National Basic Research Program of China(973 Program);The National Natural ScienceFoundation of China;The National Natural ScienceFoundation of China;The National Key Technology R&D Program of China;The National Key Technology R&D Program of China;The Fundamental Research Fundsfor the Central Universities;Electronic Development Fund of Information Industry Ministry

摘要:

当前常使用多个云共同存储数据,以保证用户数据可靠性。为减少存储成本和修复带宽,提出了一种使用功能性修复再生码(FRC)将数据编码为多个数据块,并分布于不同的云中的方法。该方法减少了多个云中单个云发生数据丢失时需要从网络上传输的数据量,并减少了修复成本,已成功地应用于所构建的云存储系统Ustor中。实验表明:与不编码比较,冗余编码给系统增加了5%~10%的响应时间开销,但可保障节点失效;FRC码编、解码和修复速度与里德-所罗门(Reed-Solomon或RS)码基本相当,256 MB大小文件编码时间差距在0.5 s以内;FRC码修复时与传统的RS码相比减少了25%以上需要下载的数据量。

关键词: 云存储, 再生码, 纠删码, 修复带宽

Abstract:

To ensure reliability of user's data, multiple clouds are responsible for storing the data. For purpose of mini-mizing storage cost and repair bandwidth, a method of utilizing functional regenerating codes(FRC) to encode data to several blocks, which are further distributed to different clouds, was proposed. In this way, amount of data downloaded through the network when data loss happens in a single cloud is cut down, and repair cost is cut down as well. This method was successfully applied in a cloud storage system Ustor. Observations can be drawn from the experimental re-sults: first, erasure coding guarantees data reliability, increasing 5%~10% response time compared with non-coding; sec-ond, speed of FRC encode, decode and repair are a little slower than Reed-Solomon(RS) code, and the time gap of en-coding 256MB by two codes is less than 0.5 seconds; third, compared with traditional RS codes, FRC reduces more than 25% data amount needed to download during a repair in the experiment.

Key words: cloud storage, regenerating codes, erasure codes, repair bandwidth

No Suggested Reading articles found!