From WiscKey to Bourbon: A Learned Index for Log-Structured Merge Trees

TL;DR

Bourbon introduces a machine learning-based LSM tree that enhances lookup speed by learning key distributions, achieving significant performance improvements over existing systems.

Contribution

It presents BOURBON, a novel LSM tree design that uses greedy piecewise linear regression and cost-benefit strategies for improved lookup efficiency.

Findings

Bourbon improves lookup performance by 1.23x to 1.78x.

It effectively learns key distributions using regression.

Experimental results validate its superiority over state-of-the-art LSMs.

Abstract

We introduce BOURBON, a log-structured merge (LSM) tree that utilizes machine learning to provide fast lookups. We base the design and implementation of BOURBON on empirically-grounded principles that we derive through careful analysis of LSM design. BOURBON employs greedy piecewise linear regression to learn key distributions, enabling fast lookup with minimal computation, and applies a cost-benefit strategy to decide when learning will be worthwhile. Through a series of experiments on both synthetic and real-world datasets, we show that BOURBON improves lookup performance by 1.23x-1.78x as compared to state-of-the-art production LSMs.