Autonomous Selection of Indexes and Materialized Views in Two Commerical Database Management Systems

The selection of the most beneficial indexes and materialized views (MVs) is a well-known issue in physical database design. The chosen set of indexes and MVs is crucial to the overall performance of the DBMS as it relates to specific workloads. Since this is such an important aspect of physical database design, database administrators (DBAs) spend a great amount of time selecting the best indexes and MVs for the purpose of improving database performance. In order to reduce the total cost of ownership, several DBMSs have tools for automating the physical design structure selection process. Both System B and System A have such tools, including the System B Advisor [15] and the System A Advisor [2]. In this paper, we present an objective assessment and comparison of both tools by defining evaluation metrics, analyzing recommendation algorithms, extensively testing both tools, and by providing a discussion of the results.