关键词: 纠删码, 多副本, 数据修复, 性能优化

Abstract: Erasure codes have the advantage of low storage overhead. However, they also have the drawbacks of long recovery time and high impact on application performance. This paper presents the computation model of the time for data recovery with erasure codes, and identifies the key factors that affect the recovery performance. Thereafter, this paper chooses the computation overhead, read/write overhead, and transmission overhead as the evaluation criterion for the recovery performance. In addition, this paper analyzes how the latest efforts in this field reduce overheads from the aspects of computation, read/write, and transmission. Finally, this paper compares existing coding schemes from the aspects of recovery performance, reliability, as well as storage overhead, and then points out the future research directions.

Key words: erasure codes, multiple replicas, data recovery, performance improvement