LESS: Logging Exploiting SnapShot

EasyChair Preprint no. 692

Abstract

The in-memory key-value store provides a persistence method to ensure data durability. The currently provided methods are either to create a snapshot file of a current dataset or to write the log of the performed command in the log file. However, the snapshot method has a risk of data loss and append only logging method cause a system failure due to an increase in log file size. To prevent excessive AOF file size growth, the in-memory key-value store provides a reconstruction method, but also a performance degradation and excessive memory usage occur. In this paper, we propose a new persistence method for effective memory usage and throughput. The new approach is called Logging Exploiting SnapShot (LESS). LESS is a method that combines the advantages of a snapshot using low memory usage and the benefits of an append only logging method that guarantees data persistence. We implemented LESS on Redis and conducted experiments. A benchmark test demonstrated that the proposed approach reduces the maximum memory usage by 57% and it is 2.7 times faster than the original approach. Overall, the experimental results showed that LESS is effective for Redis.

Keyphrases: In-memory key-value store, NoSQL, Persistence Method, recovery, Redis

@Booklet{EasyChair:692,
  author = {Hanseung Sung and Minhwa Jin and Mincheol Shin and Hongchan Roh and Wongi Choi and Sanghyun Park},
  title = {LESS: Logging Exploiting SnapShot},
  howpublished = {EasyChair Preprint no. 692},
  doi = {10.29007/ntqq},
  year = {EasyChair, 2018}}