Cost-Based Logical Transformations in Presto

This technical report discusses the current state of cost-based logical transformations in Presto and identifies its limitations. It outlines detailed design considerations for overcoming these limitations and enhancing the optimizer. The report includes examples of logical transformations that should be evaluated within a cost-based model and assesses the existing cost-based transformations in Presto. A proposal to implement the Cascades optimizer framework is presented, detailing the exploration, costing, and memoization of logical and physical transformations. Key design considerations and open questions that must be addressed during implementation are also discussed. Additionally, the report reviews existing implementations of the Cascades frameworks in other popular relational databases. The document emphasizes the need for a comprehensive approach to the optimization process, which involves generating alternative execution plans and applying transformation rules to improve efficiency.