Concurrency control (CC) algorithms are important in modern transactional databases, as they enable high performance by executing transactions concurrently while ensuring correctness. However, state-of-the-art CC algorithms struggle to perform well across diverse workloads, and most do not consider workload drifts. In this paper, we propose NeurCC, a novel learned concurrency control algorithm that achieves high performance across diverse workloads. The algorithm is quick to optimize, making it robust against dynamic workloads. It learns a function that captures a large number of design choices from existing CC algorithms. The function is implemented as an efficient in-database lookup table that maps database states to concurrency control actions. The learning process is based on a combination of Bayesian optimization and a novel graph reduction search algorithm, which converges quickly to a function that achieves high transaction throughput. We compare NeurCC against five state-of-the-art CC algorithms and show that it consistently outperforms the baselines both in transaction throughput and in optimization time.
@article{arxiv.2503.10036,
title = {Modeling Concurrency Control as a Learnable Function},
author = {Hexiang Pan and Shaofeng Cai and Tien Tuan Anh Dinh and Yuncheng Wu and Yeow Meng Chee and Gang Chen and Beng Chin Ooi},
journal= {arXiv preprint arXiv:2503.10036},
year = {2026}
}