VLDB 2015: Multi-Objective Parametric Query Optimization

Classical query optimization models the cost of a query plan by a scalar cost value. Multi-objective query optimization models the cost of a query plan by a cost vector where different vector components represent cost according to different cost metrics. Parametric query optimization models the cost of a query plan by a function mapping from a multi-dimensional parameter space to a one-dimensional cost space.

We generalize all previously proposed query optimization variants and propose multi-objective parametric query optimization where the cost of query plan is modeled by a function mapping from a multi-dimensional parameter space to a multi-dimensional cost space. In this talk, I outline where this most generic query optimization variant is required and how it can be tackled.