Great case study! 

I'd love to hear more about the steps you followed to perform the server trace when you were analyzing the script.  And also it would be helpful to hear in more detail the process you went through to measure the timing runs. 

It seems that everyone does this a little differently, so if you share it please let me know! 

JT Lovell

ditto to Andy's post. Thanks for sharing your experience.

JT,

Thanks for the kudos.  

The process for measuring the timing was quite simple.  I created a script that stored the current date/time in a variable, then executed the stored procedure, than calculated the elapsed time once the procedure ended and printed that value out.  To verify the improvement, the script was altered to capture start time, run the procedure, calculate and print elapsed time, add the new index, capture start time, execute and then print elapsed time.

The server trace was used for two purposes.  The first was to identify the bottlenecks in the stored procedure.  These were identified merely by examing the duration of T_Sql events.  The second purpose was to serve as input into the index tuning wizard.

Hope that helps,

Gordon

Thanks for the followup, and great info!

JT