February 4, 2010 at 6:47 am
Hi there,
In our Dell-Server we got several Perc6/E-Controllers which have the annoying attribute, to just allow 16 drives to be combined in an Raid1+0 (2 drives form one Raid1 then stripping as a Raid 0) if you want to combine more drive you have to use 4/8/16-combinations
In total we have 24 drive. 2 declared as global hot spares, so I have 22 discs available (not dividable through 4).
Now I thought, what the heck use twice the Raid10 with 2drives and join them as a dynamic drive, that windows offers.
So here is the question: DO or DONT. Besides of restore-when-failed-issues are there other arguments not to use dynamic drives (e.g. perfomance)
If the answer is DO NOT or PREFERABLE NOT, would this be an solution:
build the to raid10 as before an use them as 2 logical drives and then put on each logical drive 1 file for one filegroup? (I know 1/4 - 1 File for each fielgroup for CPU) .
Would the server then manage to put the data where it has enough space?
Thx for your advice
Mitch
February 5, 2010 at 1:25 am
It's a little while since I used multiple DAS but if you want a secure solution only ever use hardware raid functions. If you're setting your storage array as raid 10 there's absolutely no point in having a hot spare as you already have 100% redundancy, you might want to have a couple of spare disks on hand but to take out part of the array capacity is pointless. ( not the case if you're using raid 5 )
I wouldn't advise using any windows software raid to join hardware raid - very bad idea - your chances of recovery would be poor.
When I used multiple DAS for sql server I generally used the arrays to split the databases, backups on one array, tempdb on another, application db another, logs another etc.
If I view any prod server the number of disks I have for backup, tempdb and logs will generally exceed the number I have for a reasonable production database, so I don't see any real need to try to join two arrays.
Hope I understood what you were asking.
[font="Comic Sans MS"]The GrumpyOldDBA[/font]
www.grumpyolddba.co.uk
http://sqlblogcasts.com/blogs/grumpyolddba/
February 5, 2010 at 1:52 am
I guess you mean by DAS - Direct-attached storage?!
The global hot spares are used to fill in whenever a disc crashes. As you never know, if the mirroring disc of the just crashed disc won't crash until you got the other replaced, you minimize the time, where the second disc is "alone".
At least that is, what I think it does 🙂
Coming back to my original question:
I left out the whole details of the rest of the system. I have separated arrays for temdb, logs, master/msdn/model and so on.
Then I have discs for some smaller databases and finally I have the discs I mentioned that I want to you for a bigger DWH. Till End of 2010 I estimated a total datasize of about 800 GB and 1,3 GB of Indexes in this DWH. The discs have each have a size of 136GB and the maximum Array-size I can build (in the way I mention above) is 1080 GB.
One possiblity for example is to create to arrays with total amount of 1,3 GB.
What I need is an solution where I dont have to calculate for every table whether I should put in on array-part1 or array-part2.
If we refrain from doing dynamic drives. How does the SQL server balance between two database-file for each filegroup?
Or is there maybe a better solution?
Cheers,
Mitch
February 9, 2010 at 9:17 am
My personal preference in this situation would be to break things up with multiple files/filegroups and only use hardware RAID.
You can find a bit of information at the following link (http://sqlblogcasts.com/blogs/tonyrogerson/archive/2006/08/17/948.aspx)
February 10, 2010 at 2:18 am
Cheers Mate,
that is what I was hoping for. Nowing that the database is balancing itself between the files in an filegroup fixes the problem, so I can use to Hardware-Raids and put in a file per filegroup on each raid. 😉
Thx again
Viewing 5 posts - 1 through 4 (of 4 total)
You must be logged in to reply to this topic. Login to reply