Unanswered: Performance Meassurement with SQL Server Profiler
i am currently developing a new database design and try to compare two designs.
I am using SQL Server Profiler to meassure the time a query takes. I am using the same data (amount and values), query the same data, same indexes etc. but the query time for the same query over the same data vary way to much (e.g. the same query today is 10 times faster then yesterday)
I am using a standalone server (no one else works on them), SQL Server 2005, log the queries with Server Profiler into a table and compare the duration time.
Query time is typically considered a poor measure for tuning. It is obviously the main criterion you use to identify queries that require tuning but that doesn't make it a terribly good metric to see how your tuning is going.
Why? Because the time taken is affected by so many other factors, in particular load on the server and network.
Logical IO is a commonly used metric - this is because it is utterly hardware independent (the same query applied to exactly the same data (including structure and fragmentation) will always return the same value).
Personally, I would be looking at the output of SET STATISTICS IO, SET STATISTICS TIME and the query plans rather than looking at profiler. I would only use profiler if there were references to multi statement functions that are retrieving their own data.
Googling about SQL Server query tuning might help you get more information.