Towards Empirically Driven Query Optimization
Satyanarayana R. Valluri, Kamalakar Karlapalem, Arvind Hulgeri · 2009
Query optimization is a hard problem with exponential complexity. In this paper, we analyze the properties of query plans generated by a DBMS. Using these as plan logs, we discover the least cost plan for a new query. We developed a framework which collects statistics from the plans returned by the query optimizers and uses a distance function to select the least cost plan for a new query. The experimental results validate our approach by selecting near-optimal plans for a given set of test queries.