Better Space-Time Trade-Offs for LSM-Tree Based Key-Value Stores [pdf]

Researchers have proposed a new method for improving the space-time trade-off in LSM-tree based key-value stores. The authors, led by Niv Dayan, introduce a new data structure called 'Dostoevsky' which is designed to reduce the memory usage of LSM-trees while maintaining their performance. Dostoevsky achieves this by using a combination of buffering and lazy flushing. In experiments, the authors found that Dostoevsky outperforms existing methods in terms of both space usage and write throughput. The code for Dostoevsky is available on GitHub at https://github.com/nivdayan/dostoevsky.