BIVXDB: A Bottom Information Invert Index to Speed up the Query Performance of LSM-Tree

Key-value storage systems based on LSM-trees achieve fast write performance through out-of-place updates and a hierarchical structure. However, their range query process involves filtering and sorting SSTables at each level, resulting in significant performance overhead. To enhance range query performance, we traverse all KV pairs during the SSTable creation phase when compaction, and construct an efficient global index named BIVX with LSM-tree’s bottom level’s information. BIVX allows rapid location and reading of target files during range queries, significantly improving range query performance. Based on BIVX, we developed BIVXDB and demonstrated the effectiveness of our improvements through experimental results.