Comments on: SQL Server Data File Storage

By: mrdenny

mrdenny — Mon, 07 Mar 2011 17:57:05 +0000

When you see people talking about enough storage arrays in these forums you can start to see the trends as people talk about the configs and you can start to guess pretty accurately which vendors they purchased from.

By: rickmartinez

rickmartinez — Fri, 04 Mar 2011 20:43:04 +0000

Yes! Xiotech! how did you guess?

By: hlx

hlx — Fri, 04 Mar 2011 17:36:53 +0000

If RAID 10 shared everything means that all of the drives in the array are in a single mirrored stripe set (Xiotech?), it is unlikely that any file, filegroup, or drive configuration will perform much differently than any other. You are likely giving yourself additional management headaches for little to no performance increase. If you have the capability to do load testing, try your setup performance vs. a two file (data, log) db on the same drive and I believe you will likely notice < 5% difference. Many of the “best practices” with regards to dataindexlog separation that you read really apply to I/O subsystems with multiple I/O channels, various spindle counts and RAID configurations. These in general won’t apply to your system simply due to the fact that every I/O will use the same channel and spindles.

By: rickmartinez

rickmartinez — Thu, 03 Mar 2011 20:02:42 +0000

Log file is in its own volume, our SAN is a RAID 10 shared everything. Workload is about 900 users.

By: hlx

hlx — Thu, 03 Mar 2011 20:00:21 +0000

Need more information. Please add information about your log file(s) and RAID type on the SAN drives as well as the typical workload for each file group.