In modern cloud storage systems, data is routinely dispersed across distinct availability zones (AZ) to ensure resilience and accessibility. Erasure codes, known for their high reliability and low storage overhead, are preferred over replication. However, the data recovery procedure by erasure codes would suffer from a large amount of data transfer among multiple network-separated storage devices. Although a number of new erasure codes have been proposed to minimize the repair costs, the average repair cost remains high in multi-AZ environments due to the expensive consumption of bandwidth and high latency between availability zones. To address this problem, we propose AZ-Clay-Hitchhiker-Code (ACH-Code), a novel erasure code tailored for multi-AZ cloud storage systems. By integrating the strengths of Clay code and Hitchhiker code, ACH-Code effectively reduces data transmission across availability zones during repairs. We conduct rigorous mathematical analysis, implement the ACH plugin in Ceph, and evaluate its performance through extensive experiments. Results indicate that ACH-Code achieves a 92.5% reduction in average repair cost compared to AZ-Code and a remarkable 97.5% decrease compared to traditional erasure codes, significantly enhancing the efficiency and cost-effectiveness of data recovery in multi-AZ environments.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

ACH-Code: An Efficient Erasure Code to Reduce Average Repair Cost in Cloud Storage Systems of Multiple Availability Zones

  • Yucheng Kang,
  • Jiawei Li,
  • Chenming Chang,
  • Keqiang Li,
  • Yupeng Chen,
  • Yi Zhang

摘要

In modern cloud storage systems, data is routinely dispersed across distinct availability zones (AZ) to ensure resilience and accessibility. Erasure codes, known for their high reliability and low storage overhead, are preferred over replication. However, the data recovery procedure by erasure codes would suffer from a large amount of data transfer among multiple network-separated storage devices. Although a number of new erasure codes have been proposed to minimize the repair costs, the average repair cost remains high in multi-AZ environments due to the expensive consumption of bandwidth and high latency between availability zones. To address this problem, we propose AZ-Clay-Hitchhiker-Code (ACH-Code), a novel erasure code tailored for multi-AZ cloud storage systems. By integrating the strengths of Clay code and Hitchhiker code, ACH-Code effectively reduces data transmission across availability zones during repairs. We conduct rigorous mathematical analysis, implement the ACH plugin in Ceph, and evaluate its performance through extensive experiments. Results indicate that ACH-Code achieves a 92.5% reduction in average repair cost compared to AZ-Code and a remarkable 97.5% decrease compared to traditional erasure codes, significantly enhancing the efficiency and cost-effectiveness of data recovery in multi-AZ environments.