Question

   SELECT projectID, urlID, COUNT(1) AS totalClicks, projectPage,
   (SELECT COUNT(1)
     FROM   tblStatSessionRoutes, tblStatSessions
     WHERE  tblStatSessionRoutes.statSessionID = tblStatSessions.ID
     AND    tblStatSessions.projectID = tblAdClicks.projectID
     AND    (tblStatSessionRoutes.leftPageID = tblAdClicks.projectPage OR
           tblStatSessionRoutes.rightPageID = tblAdClicks.projectPage)) AS totalViews
   FROM   tblAdClicks
   WHERE  projectID IN (SELECT projectID FROM tblProjects WHERE userID = 5)
   GROUP  BY projectID, urlID, projectPage
   ORDER  BY CASE projectID
          WHEN 170 THEN
           1
          ELSE
           0
      END, projectID

This is by no means an especially complex query, but because the database is normalised to a good level, and we are dealing with a significant amount of data, this query can be quite slow for the user.

Does anyone have tips on how to improve the speed of it? If I strategically denormalise parts of the database would this help? Will running it in a stored proc offer significant improvements?

The way I handle the data is efficient in my code, the bottleneck really is with this query.

Thanks!

Answer 1

De-normalising your database should be a last resort since (to choose just one reason) you don’t want to encourage data inconsistencies which de-normalisation will allow.

First thing is to see if you can get some clues from the query execution plan. It could be, for example, that your sub-selects are costing too much, and would be better done first into temp tables which you then JOIN in your main query.

Also, if you see lots of table-scans, you could benefit from improved indexes.

If you haven’t already, you should spend a few minutes re-formatting your query for readability. It’s amazing how often the obvious optimisation will jump out at you while doing this.