Query optimization is one of the most challenging aspects of database systems research. A database’s optimizer must account for factors like indexing, join strategies, caching, and hardware resources while navigating cost models that may not always accurately predict real-world performance. Additionally, modern databases must handle distributed and cloud-based environments, where data locality, network latency, and resource contention further complicate optimization. The problem is inherently NP-hard, requiring heuristics, approximations, and machine learning techniques to achieve near-optimal performance within practical time constraints.

The optd project seeks to develop a high-performance, extensible optimizer-as-a-service. It serves as a research vehicle for several other sub-projects, including cardinality estimation, adaptive query planning, AI-enhanced planning, and parallel execution. Our plan is to develop a query optimizer that rivals state-of-the-art implementations from commercial systems and the Germans.

People

• Alexis Schlomer
• Yuchen Liang
• Connor Tsui
• Sarvesh Tandon
• José Orlando Pereira (University of Minho)
• Jignesh Patel
• Andy Pavlo