An Optimization of Key-Value Store Based on Segmented LSM-Tree

Storage Engine is the core of the storage system, R/W performance (read and write performance) of the storage system depends on the performance of the storage engine. sLSM-Tree structure (LSM-Tree structure based on the segmented index) is proposed, which is based on the structure of LevelDB. Segmented index structure is introduced to solve the collisions brought by adding hash storage RAM index structure to the index structure parts of LSM-Tree, i.e. trie index and hash index segmentally. By this way, index speed is improved and the pressure of updating index terms by compacting is reduced. The contrast experiment was conducted about the novel segmented index method presented in this paper. From the analysis of experimental results, sLSM-Tree has a significant performance in the RAM index and R/W operation on the hard disk compared with LevelDB which uses conventional LSM-Tree storage engine.