SQL Server Storage for multiple data files

Question

Recently, our company wants to add 4TB on the drive size for SQL Server. We have two drives now which have different size and data files. In order to make sure our SQL Server has the best performance, where should I add those space:

G: Drive: 19TB (16 Data files) H: Drive: 3TB (3 Data files)

Should I add the space on H drive and create more data files in it? Any suggestions?

Answer 1

My Data files are always kept in a separate drive from where the SQL install took place. Therefore, databases like master, model, msdb, and tempdb are always on one drive, while my (user) databases are on (one or more) separate drives.

This is not so much for performance reasons as it is for Disaster Recovery/Security. it is much easier to restore a SQL Instance install than it is to recover the data.

If your SQL install is on either of these drives, I would chose the opposite one. If your SQL install is on neither, then I'd probably chose to add additional space to the 19TB drive (assuming it has more free space than the 3TB drive)

Answer 2

Space will not be the only factor in where you place the new space. If you are talking about expanding the capacity of the existing drives you should look at what do the underlying volumes have backing them. I.e how many physical drives are assigned to the drive arrays and what the RAID levels are for those arrays.

For HDDs the more drives you have in an array the better it will perform. The amount of improvement will be a "Your mileage will vary" answer based on what your hardware is and how it is configured.

As an example of how to start at making that determination let assume the following:

Drive G has 20 1TB drives in a RAID 10 configuration. Drive H has 22 146GB drives in a RAID 10.

Your new drives are another 28 146GB drives. You might want to add those to the H drive array as they are the same drive spec.

If your new drives are not the same to the drives already being used in one of the two arrays I would do some capacity planning to see how to break down the new drives into multiple arrays and spread some of the existing data files across things based on the access pattern of your data.

You can also check out some posts from Brent Ozar and team on disk io

For SSD disks there is a completely different way of looking at things. Check out the video from Brent