Challenges in building a DBMS Resource Advisor

Administration increasingly dominates the total cost of ownership of database management systems. A key task, and a very difficult one for an administrator, is to justify upgrades of CPU, memory and storage resources with quantitative predictions of the expected improvement in workload performance. We present a design and prototype implementation of a Resource Advisor that is able to answer “whatif” questions about DBMS performance under hypothetical conditions. We discuss the design issues and challenges involved in building such a Resource Advisor, as well as our experiences in building a prototype Resource Advisor for SQL Server.