Sprinkling Selections over Join DAGs for Efficient Query Optimization

TL;DR

This paper introduces a novel query optimization method that uses join graphs and DAGs to efficiently generate optimal join orders, improving performance over traditional AND/OR DAG strategies especially with complex queries.

Contribution

It proposes a new approach combining join graphs and DAGs for faster, scalable query optimization and demonstrates its effectiveness through experiments on TPC-D/H benchmarks.

Findings

Significant reduction in optimization time compared to traditional methods

Improved query execution plans for complex join queries

Effective in multiple query optimization and view selection scenarios

Abstract

In optimizing queries, solutions based on AND/OR DAG can generate all possible join orderings and select placements before searching for optimal query execution strategy. But as the number of joins and selection conditions increase, the space and time complexity to generate optimal query plan increases exponentially. In this paper, we use join graph for a relational database schema to either pre-compute all possible join orderings that can be executed and store it as a join DAG or, extract joins in the queries to incrementally build a history join DAG as and when the queries are executed. The select conditions in the queries are appropriately placed in the retrieved join DAG (or, history join DAG) to generate optimal query execution strategy. We experimentally evaluate our query optimization technique on TPC-D/H query sets to show their effectiveness over AND/OR DAG query optimization strategy. Finally, we illustrate how our technique can be used for efficient multiple query optimization and selection of materialized views in data warehousing environments.