Hybrid Query Optimization with Traditional and Learning-Based Optimizers

Traditional cost-based query optimizers are highly effective for many database workloads, especially for short and moderately complex queries. However, for queries involving many joins and large search spaces, optimization can become expensive and inaccurate. Recent learning-based query optimizers have shown promising results for complex queries, but they may introduce unnecessary overhead for simpler queries. This project aims to develop a hybrid query optimizer that combines the advantages of both approaches. The main idea is to analyze an incoming query and dynamically decide whether it should be handled by a traditional cost-based optimizer or a learning-based optimizer. A simple initial approach may use query complexity indicators such as the number of joins, while more advanced solutions may consider additional features or employ a lightweight learned classifier. The project will investigate questions such as: When should a traditional optimizer be preferred over a learned optimizer? Which query characteristics are useful for making this decision? How to build a learned query optimizer dedicated for long and complicated queries? Can a hybrid approach improve overall workload performance compared with using either optimizer alone?

Students will implement a prototype and evaluate it on benchmark workloads containing queries of different complexities. The evaluation will consider query execution time, optimization time, and total end-to-end latency.

The project is suitable for students interested in database systems, query optimization, machine learning for systems, and experimental performance evaluation. Basic knowledge of databases and programming is expected; prior machine learning experience is helpful but not required.