Database system performance problems have a cascading effect into all aspects of an enterprise application. Database vendors and application developers provide guidelines, best practices and even initial database settings for good
performance. But database performance tuning is not a one-off task. Database administrators have to keep a constant eye on the database performance as the tuning work carried out earlier could be invalidated due to multitude of reasons. Before engaging in a performance tuning endeavor, a database administrator must prioritize which tuning tasks to carry out first. This prioritization is done based on which tuning action would yield highest performance benefit. However, this prediction may not always be accurate. Experiment-based performance tuning methodologies have been introduced as an alternative to prediction-based performance tuning approaches. Experimenting on a representative system similar to the production one allows a database administrator to accurately gauge the performance gain for a particular tuning task. In this paper we propose a novel approach to experiment-based performance tuning with the use of a context-aware application model. Using a proof-of-concept implementation we show how it could be used to automate the detection of performance changes, experiment creation and evaluate the performance tuning outcomes for mixed workload types through database configuration parameter changes.