Semi-materialization: a technique for optimizing frequently executed queries

摘要

We propose a technique called semi-materialization to efficiently implement frequently executed queries in the context of relational database systems. The method is based on storing redundant data that represents an intermediate stage of query evaluation. The redundant data is used to efficiently construct the result of the query, yet is easy to maintain. We then compare the performance of our approach with two other approaches commonly used to implement frequently executed queries, full-materialization and query modification. This is done analytically, and verified by simulation. The results indicate that the method is beneficial not only for complex view definitions, but also for simpler select-project-join expressions under certain operating conditions.