Abstract

The Cascades Query Optimizer Framework is a tool to help the database implementor (DBI) in constructing a query optimizer for a DBMS. It is data model independent and allows to code a query optimizer by providing the implementations of the subclasses of prede ned interface classes. When the implementations of the required classes are provided properly, the generated optimizer produces the optimum execution plans for the queries. Although providing the complete set of rules and thus nding the optimum execution plans are bene cial for most of the queries, the query optimization time increases unacceptably for certain types of queries, e.g., for star queries. Hence it is important to be able to limit the number of alternative plans considered by the optimizer for speci c types of queries by using the proper heuristics for each type. This leads to the concept of region based query optimization, where di erent types of queries are optimized by using di erent search strategies in each region. This paper describes our experiences in developing a region based query optimizer through Cascades. Cascades' guidance structures provide the facilities required for the design and implementation of a region based optimizer. The performance comparisons between a region based query optimizer and an exhaustive (which uses the complete rule set without heuristic guidance) query optimizer, both generated through Cascades, indicate that the region based optimizer has a superior performance. In the performance analysis, both the sum of optimization and execution times, namely the response time, and the quality of the plans generated are investigated. 1

Citations