T-SQL Tuesday #040: File and Filegroup Wisdom
It’s that time again for the T-SQL Tuesday party! This party was created by none other than Adam Machanic (Twitter). If you are interested in hosting a party at some point this year give him a shout; you need to have participated in two T-SQL Tuesdays along the way and also maintain your own blog for at least 6 months.
Now that we have what the party is all about let’s get into what this month’s party is centered around Filegroups and his hosted by Jen McCown / MidnightDBA
My focus today is garnered toward indexes on filegroups and what they can do to your index strategy. I’m a big fan of having strategies when tackling issues, problems, or even believe it or not from the beginning of a project. Placing indexes on filegroups carefully can improve query performance (at the same time I want to note that indexes can also hurt performance in some situations so thorough testing needs to be taken into consideration).
Back from my 2008 R2 studies, if my memory serves me correctly, indexes are stored in the same filegroup by default; a non-partitioned clustered index and the associated table always reside in the same filegroup however you can do one of three things:
- You can partition both clustered and non clustered indexes to span multiple filegroups
- Create non clustered indexes on a filegroup
- Move a table from one filegroup to another
You can achieve performance gains by created non clustered indexes on a different filegroup if the filegroups are using different physical drives. The data and index information can be read in parallel by the multiple disk heads when the physical drives are on their own controllers.
One cannot foresee the access that will transpire or when it will happen, a better decision to spread your tables and indexes across all file groups might be of help. This would guarantee all disks are being used and accessed because all data and indexes would be spread evenly across all disks.
To bring this all back together you can think of a filegroup in its simplest of forms. Every database that you create has at least a data file and a log file, and every database has a primary filegroup. The filegroup contains the primary data file and any secondary files that are associated with it. One filegroup can contain multiple mdf/ndf files.
In the end I have seen significant gains with indexes being placed on specific filegroups, but as I stated before it is good to test all this out. Set up some scenarios on your test server and start doing some test cases to prove different theories and ideologies. One thing to remember as well is not every case is the same; ensure that the decisions you are making is good for what you are working on; never take a suggestion and drop it into a production environment. Prove the statement to be true or false no matter who it comes from.
Well, that’s a wrap for today’s party. Until next month…….