Analyzing Wait Statistics
One of my main recommendations for troubleshooting a performance problem when a particular issue is lacking information is to run a SQL trace or extended events session to find out the most expensive queries. This method will usually show us the most expensive queries based on one or more metrics including duration, CPU usage, logical and physical reads, writes, rows, and so on. Many times the metric of choice is duration, or how long the query took to execute. We usually assume this entire duration is usable work, which of course can be optimized. In a lot of cases, this method can find the troubled queries, and tuning them solves the performance problem.