I have a select store procedure which cannot be change, so I tried to add index in the tables to get a better performance. My questions is which one is faster? Should I look at the duration? or Reads? If duration then the 2nd one is faster, if Reads, the last one is faster. Thanks, Terrence Here is the result form TRACE. Duration CPU Reads Writes 175 32 240 0 62 47 134 0 after change TABLE1, one index to cluster-index 44 31 136 0 after change TABLE1, all index to cluster-index 123 62 27 0 after change TABLE2, all index to cluster-index
I would look at the reads and then duration. -- Frank Kalis Microsoft SQL Server MVP http://www.insidesql.de Ich unterstütze PASS Deutschland e.V. http://www.sqlpass.de)
...with the help of Execution plan from QA. Satya SKJ Moderator http://www.SQL-Server-Performance.Com/forum This posting is provided “AS IS†with no rights for the sake of knowledge sharing.
reads cannot be compared from one operation to another, ie, one execution plan to another. presumably the index mods changes the execution plan. so be sure to run each condition twice, take the results from the second look at duration for user response. cpu for system thruput note that the Profiler reports cpu and duration in approx multiples of 10ms for 1 CPU, and 15.625 for multi-proc systems, go the sqldev.net for a fix to allow profiler to have 1ms time resolution