October 30, 2013 at 8:34 am
Hi I have a new install of SQL on a server spec'd by a former coworker
A Rac server with a terra drive which is local and partitioned into 2 (100GB for OS and 800+ for data)
the server is connected to our compellent with 6 LUNs assigned as below
DB drive 1TB
TemDB 100GB
LOG 500GB
TempDB LOG 100GB
Backup 1TB
Service drive 200GB
I want to know. when I right click on databases and select restore from device, I click add and it takes a long time before anything is displayed +- 2 mins is this because of the amounts of drives or just something in the registry?
October 30, 2013 at 11:32 am
That time to restore is based on getting the file lists from the drives. The time seems to vary, but I'm not sure the number of drives attached is the problem. It could be, but the server should read the default path from the registry and then load that drive's file structure. I'd think it only loads the drive letters from other files, but I could be wrong.
October 30, 2013 at 11:38 am
Sl33py (10/30/2013)
I want to know. when I right click on databases and select restore from device, I click add and it takes a long time before anything is displayed +- 2 mins is this because of the amounts of drives or just something in the registry?
Are you doing this local on the server or from your desktop (other remote computer)?
Shawn Melton
Twitter: @wsmelton
Blog: wsmelton.github.com
Github: wsmelton
October 30, 2013 at 12:37 pm
In general, more disks can improve performance, not generally hurt it. But, you also have to have more controllers to really achieve meaningful benefits. In general, based on what you're saying, I would not assume that you should see performance degradation from just those few disks. I'd suspect the problem is elsewhere.
"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt
Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning
November 1, 2013 at 11:24 am
You mentioned a Compellent, which I presume to be a Compellent SAN. If you really want to dig into this, you'll need to work with the SAN team to map (using whatever terminology Compellent uses):
OS drive letters/mount points to LUNs
LUNs that are part of larger sets of LUNs (metaluns)
LUNs to RAIDsets (RAIDset: a collection of sets of storage bound together with a given RAID level)
RAIDsets to spindles (or SSD units)
And then you go back up the chain:
spindles to every RAIDset using each spindle
RAIDsets to LUNs
LUNs that are part of larger sets of LUNs (metaluns)
LUNs (metaluns) to the OS drive letters/mount points
After which exercise, you'll be able to see what your storage is shared with. Perhaps your LUNs each have spindles dedicated to them that do nothing else. Perhaps every single LUN you have is part of a storage pool that's shared with the Exchange email server, so heavy email traffic can slow down your database. Perhaps one or two of your LUNs is shared with a data warehousing box.
If you want to dig into storage performance, you must know the entire layout as it pertains to your system - otherwise, you can easily end up chasing phantoms caused by resources shared with other applications that have sporadic load cycles. Bottlenecks on the fiber channel connections are also possible, though less common.
November 1, 2013 at 12:12 pm
I'd like to add on and ask a question as well, when you click "Restore from Device" does the application actually go through all your network adapters one by one and check to confirm that there is something on the back end of each connection? Maybe network or back end device could be an issue if some devices don't reply in a timely fashion??
November 3, 2013 at 10:10 pm
Morning All
Thank you all very much, The issue we were having, it seemed that this server was a broken cluster and was reformatted. The issue was was that the old Quorum drive was still attached to this config and errors were generated on the compellant for the old cluster name.
We removed the drive and it is fast again.
Viewing 7 posts - 1 through 6 (of 6 total)
You must be logged in to reply to this topic. Login to reply