Conference Schedule - PGCon 2020

Introducing LSM-tree into PostgreSQL, making it as a data gobbler

Date: 2020-05-28
Time: 14:00 - 14:45
Room: Stream 1
Level: Intermediate

Data storage engine is always the key to the performance of a database system. Traditionally, storage engine is implemented based on B+-tree or B-tree. Compared to traditional B+tree or B-tree, LSM-tree as another important data structure, recently attracts great attention from developers due to its wide usage in many open source projects including in web browsers and many data processing systems. There are some well-known incarnation of LSM-tree, such as LevelDB and RocksDB to make the index practical in the industry. Now, we bring it to the world of PostgreSQL. Users can utilize LSM-tree via Postgres‘ FDW through our implementation.

So in the talk, we will cover these contents: 1. The motivation of introducing LSM-tree 2. How does LSM-tree work? 3. How does the specific implementation of LSM-tree work, such as LevelDB and RocksDB? What is their architecture? 4. How do we incorporate these existing LSM-tree implementations? 5. Some experiment results

The project is open source and available here:



The following slides have been made available for this session:


Shichao Jin