WOKV: A write-optimized key-value store

In big data era, LevelDB is widely deployed in enterprise server for massive storage. However, with the growing size of stored data, LevelDB will inevidablely frequently call the background compaction thread to compact data between level, resulting in write amplifications and therefore degrading performance. In this paper, we propose WOKV, a Write-Optimized Key-Value store, to mitigate the LevelDB compact write amplification and improve system performance. WOKV adopts a Least-Rewrite Compaction Strategy to mitigate the write amplification induced by compaction operations, and uses Multi-Thread and Multi-Buffer to improve system performance. Evaluation results show that WOKV outperforms LevelDB significantly.