Automatic performance diagnosis for changing workloads in DBMSs

Database performance is directly linked to database management system (DBMS) resource allocation. Complex relationships between DBMS resources make problem diagnosis and performance tuning difficult, time-consuming tasks. Database administrators are currently required to retune the DBMS as databases grow and workloads change. Performance can be increased and cost of ownership reduced by automating the tuning process, starting specifically with the diagnosis of resource allocation problems. In this paper, we overview our automatic diagnosis framework designed to determine resource problems. We present our results demonstrating the ability to correctly identify system bottlenecks for a generic on-line transaction processing workload when new transactions are added to the workload.