Automated Diagnosis and Control of DBMS Resources

Traditional methods of allocating resources for a DataBase Management System (DBMS) include the manual determination and allocation of each resource. Database resources are commonly tuned to achieve peak performance for a workload that is well-known and predictable. As databases become more commonly used in embedded systems and on the Internet, it is impossible to predict the workload that the DBMS will have to handle. From On-Line Transaction Processing (OLTP) to OnLine Analytical Processing (OLAP), DBMSs must be able to provide peak performance for not only different types of workloads, but multiple workloads on a single system. There is a clear need for a system that can determine what resource is causing poor performance and how to fix the problem.