Dept. of Computer Science and Engineering
Oregon Graduate Institute of Science & Technology
Join-order optimization plays a central role in the processing of relational database queries. This dissertation presents two new algorithms for join-order optimization: a deterministic, exhaustive-search algorithm, and a stochastic algorithm that is based on the deterministic one. The deterministic algorithm achieves new complexity bounds for exhaustive search in join-order optimization; and in timing tests, both algorithms are shown to run many times faster than their predecessors. In addition, these new, fast algorithms search a larger space of join orders than is customary in join-order optimization. Not only do they consider all the so-called bushy join orders, rather than just the left-deep ones, but-what is more unusual-they also consider all join orders that contain Cartesian products. The novel construction of these algorithms enables them to search a space including Cartesian products without paying the performance penalty that is conventionally associated with such a search.
Vance, Bennet, "Join-order optimization with Cartesian products" (1998). Scholar Archive. 177.