Author(s)

Xiang Liu

Corresponding Author

Xiang Liu

Abstract

The advent of the era of big data has put forward new requirements for data management, and writer-intensive applications are appearing more and more frequently, which makes the LSM-Tree based key and value storage system play an increasingly important role in today's information society.In order to optimize the writing-intensive data structure, LSM-Tree takes full advantage of the fact that the sequential read/write performance of the disk is much higher than that of random read/write performance, and accumulates the data written to the disk in batches.However, this design does not take into account the characteristics of the data itself. In this paper, for the workload with spatial locality, the prefix heat statistics method is used to accurately identify the areas with high access frequency, and further the cold and hot separation method is used for targeted management, which effectively improves the system performance.At the same time, in order to avoid excessive memory consumption caused by statistical data heat, this paper adopts an adaptive method to prune and split the statistical tree, effectively reducing the cost of this part.

Keywords

KV store, Lsm-Tree, Separation of heat and cold, Prefix tree, Access statistics