July 13, 2011 at 8:47 am
Perry Whittle (7/12/2011)
RAID10 can offer fast reads depending upon the controller used. Modern RAID10 arrays will read from all online disks to improve performance
Good point Perry, and one which most aren't aware of (yet). The ability to read down both sides of the disk set can provide much improved read performance from RAID10.
Best,
Kevin G. Boles
SQL Server Consultant
SQL MVP 2007-2012
TheSQLGuru on googles mail service
April 24, 2012 at 7:00 am
Hi peeps,
Did anyone ever do a comparative test on the recommended solutions? We are faced with exactly the same situation at the moment. We have 8x300GB drives and require a minimum of 350GB for data files based on historical growth patterns. So I too proposed a three RAID setup of:
DriveRAIDCntSizeDescription
C12280OS, Software, Backups
D104560Data Files (mdf, ndf)
E12280Log Files (ldf), System Databases
Now people are concerned that it wont perform any good as the tempdb shares resources with the data and log files. I agree with this statement and would have loved to combat this with two additional drives configured in RAID 1 but this isn't possible as the server only supports eight drives. So I believe that this is a much better performing solution than:
DriveRAIDCntSizeDescription
C12280OS, Software, Backups
D54840Data Files (mdf, ndf)
E12280Log Files (ldf), System Databases
How can we perform tests, would SQLIO yield any conclusive results?
April 24, 2012 at 7:25 am
Orion,
It depends.
It depends on your virtual io stats and which databases and files it reveals are most busy. (tempdb, user data files, log file.)
It depends on if the busy file activity is read or write mostly.
After you know that, you can run sqlio, or better yet, IOMeter to see how the drives perform compared to each other. And choose your raid type and locate the right files based on what is needed.
Jim Murphy
http://www.sqlwatchmen.com
@SQLMurph
April 24, 2012 at 7:43 am
Orion Pax (4/24/2012)
So I too proposed a three RAID setup of:
Drive RAID Cnt Size DescriptionC 1 2 280 OS, Software, Backups
D 10 4 560 Data Files (mdf, ndf)
E 1 2 280 Log Files (ldf), System Databases
You could put Tempdb out on E, backups and data on D and OS, software and Logs on C. There are many other configurations you could consider but you need to think about disk layouts. You have no hot spare and assigning TEMPDB to a 280GB drive would possibly be wasting valuable space 😉
Orion Pax (4/24/2012)
Now people are concerned that it wont perform any good as the tempdb shares resources with the data and log files.
With the disk configurations available to you, you are going to be limited and this should be realised from the start!
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
April 24, 2012 at 7:48 am
in the original suggestion it was stated to use a raid 1 (mirror) for the C and D drives - but this would mean your d: drive that takes the backup is smaller than your data drive.......
how about as an alternative
RAID 5 with 4 disks for C: and D: (O/S and backup)
RAID 1 with 2 disks for data
RAID 1 with 2 disks for log
MVDBA
April 24, 2012 at 8:24 am
Thank you for all the feedback.
Jim, it seems that the system on average is 60% writes and 40% reads.
Perry, the tempdb will be located on E as per the original grid as it is a system database.
Michael, as mentioned in my original response we require 350GB of space and RAID 1 would only yield +-280GB of usable disk space on a 300GB drive. Backups will be significantly smaller as a third party compression tool (Idera, sorry RedGate) will be used so not a significant amount of space will be required for the backups.
April 25, 2012 at 9:15 am
This is a one-year-old thread. Please start new threads, even if they are similar topics.
Best,
Kevin G. Boles
SQL Server Consultant
SQL MVP 2007-2012
TheSQLGuru on googles mail service
Viewing 7 posts - 31 through 36 (of 36 total)
You must be logged in to reply to this topic. Login to reply