Efficient B-Tree Insertions Using Proximal Policy Optimization and Hierarchical Attention Models

Alexander Kastius, Nick Lechtenbörger, Felix Schulz, Johann Schulze Tast, Rainer Schlosser, Ralf Herbrich

Keywords: Reinforcement Learning, Optimization, Databases, Attention, B-Trees

TL;DR: A hierarchical attention-based model is presented with the goal of optimizing operation scheduling on dynamically managed B-trees.

Abstract: B-trees are a fundamental component of any large database management system. They can grow to noticeable sizes, but their handling is non-trivial. We present a novel approach to use attention-based models with weight sharing across the hierarchical structure of the tree to parse such large trees fast and without the need for excessive training on large clusters. We present a use case in which the model is used in conjunction with PPO to manage write operations on such a tree.

Submission Number: 24