Cost-Based Query Optimization via AI Planning

Authors: Nathan Robinson, Sheila McIlraith, David Toman

AAAI 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results demonstrate the effectiveness of AI planning techniques for query plan generation and optimization.
Researcher Affiliation Academia Nathan Robinson College of Engineering and Computer Science Australian National University Canberra, Australia Sheila A. Mc Ilraith Dept. of Computer Science University of Toronto Toronto, Canada David Toman D.R. Cheriton School of Computer Science University of Waterloo Waterloo, Canada
Pseudocode Yes Algorithm 1 DF
Open Source Code No The paper does not provide any explicit statements about releasing source code or links to a code repository.
Open Datasets No We tested our planning systems on randomly generated database schemata and queries.
Dataset Splits No The paper describes using 'randomly generated database schemata and queries' but does not specify explicit training, validation, or test dataset splits, percentages, or methodology for splitting, nor does it reference predefined splits.
Hardware Specification Yes All experiments were run on a 2.6GHz Six-Core AMD Opteron(tm) Processor with 2GB of memory per experiment.
Software Dependencies No Our Python implementation of DF, and 'The code is based on the eager search algorithm in Fast Downward (Helmert 2006).' (No specific version numbers for Python or Fast Downward are provided).
Experiment Setup Yes Each generated schema consists of tables with between 2 and 10 attributes. Each table has a random size of between 10k and 500k tuples and 200 tuples are assumed to fit into a page of memory. ... An upper bound B on plan cost was produced by running DF for 5 seconds and then A* and GR were run with the initial bound B with a time limit of 30 minutes. ... Each algorithms was run for a total time of 2 minutes and plans were recorded as they were produced.