High Level Performance Estimate of Relational Queries

Performance estimate of queries is a necessary part of any efficient database design methodology. In this paper, a high level performance estimator for relational queries is presented which is different from conventional evaluators such as SYSTEM R [Selinger 1979] in that performance is predicted without the details of the low level constructs such as links and indices. Rather, abstractions and reasonable assumptions of these low level constructs are used in a set of formulas to estimate the performance of a set of queries against a schema. The major results of the paper are (1) The realization and motivation of the need of high level performance estimators of this kind, and (2) a fast way to estimate the cost of N-way joins. The second result is interesting in that the algorithm for N-way joins is found to be similar to matrix multiplication optimization, with straight forward This work was supported by the Applied Mathematical Sciences Research Program of the Office of Energy Research U.S Department of Energy under contract W-7405-ENG-48.