tools for profiling and tuning application code remain disconnected from the profiling and tuning tools for relational DBMSs. This makes it challenging for developers of database applications to profile, tune and debug their applications, for example, identifying application code that causes deadlocks in the server. We have developed an infrastructure that simultaneously captures both the application context as well as the database context, thereby enabling a rich class of tuning, profiling and debugging tasks that is not possible today. We have built a tool using this infrastructure that enables developers to seamlessly profile, tune and debug ADO.NET applications over Microsoft SQL Server by taking advantage of information across the application and database contexts. We describe and evaluate several tasks that can be accomplished using this tool.
Surajit Chaudhuri, Vivek R. Narasayya, Manoj Syama