Sql-server – SQL Server 2000 optimzation job hasn’t run for a while

jobsoptimizationsql-server-2000

Our SQL Server 2000 used to run a Sunday morning 3am job of optimizing the database. Somehow it was turned off and hasn't run in a year. We have noticed slow access to the database.

We turned it back on and it will run for about 4 hours and complete. Our issue is when it runs on its scheduled Sunday at 3am it will cause the applications that connect to stop working as long as it runs. Before it was shut off the application would continue to work but I believe since it has been off for a year and the job only ran once it wasn't enough to clean the DB up enough.

Should we run it three times in a row? Or what is the best way to fix this?

thanks in advance!

Best Answer

A lot can have happened in a year to cause performance issues. The fact that your "clean up job" did not run might or might not be the primary reason.

Before you start optimizing performance, the first thing you need to do is finding out what your biggest problem is and then address that. (Independent of the SQL Server version, or really independent of the system you are working with.)

SQL Server 2000 provides a limited set of tools you can use for this investigation. Start by tracing batch and procedure executions to see which of them take the longest and do the most reads. It is not unlikely at all that over the year the data changed in a way that a new index is now required.

If in the end you come to the conclusion that the indexes in place are the best for your workload, you can look at this "clean up job" again.

First you need to figure out (on a per table basis) if a reorganization or a rebuild is required. Reorganization usually is faster but has a smaller improvement impact. after a year of "neglect" you most likely need to execute a rebuild on all tables. See http://technet.microsoft.com/en-us/library/cc966523.aspx for more details, particularly the DBCC DBREINDEX vs. DBCC INDEXDEFRAG section.

As Shawn mentioned, each table rebuild/reorganize action requires a table lock in SQL Server 2000. To mitigate that there are two options:

  1. Upgrade to a more modern version of SQL Server that supports online index maintenance

  2. Clean up one table at a time and then take a significant break to allow other processes to be worked on. Depending on how many tables you have do one a night, or a few with an hour break in between each. In the following nights address other tables. You could also, each night address the 3 - 5 most fragmented tables. DBCC SHOWCONTIG can show you how fragmented a table is. (See above link for details.)