September 9, 2003 at 11:20 am
Hi all-
We are getting RAID5 for our new database server. And one of the arrays contains 5 disks 32GB each. Is it possible to place the data/log/indexes on seperate disks within the array? If yes, I greatly appreciate if you can direct to any information on the NET that talks about this.
Thanks for your time
September 9, 2003 at 11:45 am
Depending on the database size, you may not see the different by placing mdf and ldf in same RAID set if the database is small. But it is alway be good practice to separate the mdf and ldf in different RAID set, especially placing ldf in RAID 1 or RAID 10.
September 9, 2003 at 12:02 pm
Couple of months back, when we increased disk space on our production server, we increased by 2*72 gig with RAID1 (with mirroring) as a separate array apart from our existing 5*36 gig RAID5 (with parity).
After adding the disk space, I started using RAID 1 for storing the backups (for both Database backups & Log backups). I could see much reduction in decreasing the time it takes to backup database backups. It reduced from 47 mins to 8-10 mins. Now by placing like this we can only see the efficiency in the database maintenance.
Now I am planning to move my logs (.LDF) to RAID 1 to increase efficiency in Database Server performance.
Here is what I am planning:
RAID 5 : Database File (.MDF) & Tran. Log File Backups (.TRN)
RAID 1 : Database Log Files (.LDF) & Database Backups (.BAK)
Any thoughts are welcome!!!
.
September 10, 2003 at 5:03 am
Interesting. You're gaining some speed because you're reading from one and writing to the other of course. As long as they are truly separate disks that will work, not if you've partitioned drives to just look like RAID5/1.
Andy
September 10, 2003 at 7:09 am
I like having my data files on a different drive than my backups (full and transaction logs).
Patrick T Ward
September 10, 2003 at 7:19 am
Someone can give us what price difference between RAID 1 and RAID 5?
Thanks
feifei liu
Rachel
September 10, 2003 at 7:43 am
I know I'm not answering the question, but it isn't entirely off topic.. I just wanted to remark that we had substantial problems with performance while using RAID 5 (almost everything was slow, but CPU usage was mostly below 50%) and decided to switch to RAID 1 0. It helped a lot, especially when we placed tempdb on a separate disk without any mirroring or RAID or anything, and used smaller (18GB) and quicker disks for the RAID. In connection with daily full backup and log backups done every 15 minutes we consider it secure enough.
Of course the best solution depends on amount of inserts, updates and selects performed on the DB, as well as other issues - but our particular DB didn't perform well with RAID5. BTW, our production DB is now about 90 GB (original size 2 years ago was 25 GB).
September 10, 2003 at 8:02 am
I agree with vladan 100%. We have ~ 300GB database on a server with 6 X 144 GB drives in RAID 5. I think it was done this way to save some $$ on disks, but let me tell you , it is not worth it. The disks are saturated all the time (with lots of disk waits), with 2 GB memeory left idle and the processors barely reach 20%. The disk queue is often in 30s and reaches 150. So I suggest if you have large databases (50 GB +) I would stay away from RAID 5. You may save a few bucks on the disks, but you are wasting a whole lot of other server resources. BTW, I am pushing the mgmt to buy some SAN storage. I think for databases of this size internal drives are not gonna cut it.
Regards.
September 10, 2003 at 8:03 am
quote:
Someone can give us what price difference between RAID 1 and RAID 5?
Strictly speaking Raid 1 is about 33% more expensive than RAID 5. This is because mirroring is 1 to 1. With striping you get 2 to 1 because the parity is written to the 3rd device.
Joe Johnson
Joe Johnson
NETDIO,LLC.
September 10, 2003 at 8:10 am
quote:
Someone can give us what price difference between RAID 1 and RAID 5?
The main price will be in the hard drives and any extra equipment required to contain these drives.
A good reference is:
http://www.pcguide.com/ref/hdd/perf/raid/levels/
Thomas Rushton
blog: https://thelonedba.wordpress.com
September 10, 2003 at 10:57 am
Thanks a lot for your responses. Our infrastrucutre team made a new decision today that they would be getting SAN for all the servers (including my DB server). And I was told that I don't have to worry about where to place my data/index and log files. SAN does the magic!!! Is this accurate?
Thanks for your time in advance
September 10, 2003 at 11:02 am
I am new to SAN. Can someone shed some light on this? It is my understanding that its a high end fileshare.
.
September 10, 2003 at 11:42 am
We have used SANS from EMC (Symmetrix) before for our databases. They have a custom RAID configuration RAID S. But on the database front you would still have to separate the data, log and temp db. What I mean is you do not want to take one big drive and put everything on it. We would request for a 250 GB data drive, 100 GB log drive and 100 GB tempdb drive. We used direct tape backups using TSM or veritas. But the performance is definitely better. Since the SANs are scalable, you can add additional drives to the DB server as required on the fly.
Hope it helps.
September 10, 2003 at 1:35 pm
Not exactly high end fileshares. There are several other posts about SAN's, might want to look at them. SAN's are typically highly available, high performance, and extendable. You still want to give some thought to how you divide up your space (and take into account what limitations your SAN places on you), but file placement is not nearly as critical because the SAN will have anywhere from 512m to 2g of cache. As long as you dont outrun the cache, disk time doesn't matter. You also get features like snapshots and long distance replication.
Andy
September 11, 2003 at 8:28 am
This thread could not have come along at a more apt time.
Situation
Our Production Sql server is exhibiting Network Locks and pageiolatch_sh locks.
the configuration of the diskc is a single RAID 5 array of 36GB drives.
The Processors are running at 15% and the memory (2GB) is running about 10%
The Disks are running @ 100% !!!
My proposed solution is as follows
1) leave the database files on the RAID 5 array
2) Add a second RAID 1 array
3) Add a Single disk for the TempDB
RAID 5 : Database File (.MDF) & Tran. Log File Backups (.TRN)
RAID 1 : Database Log Files (.LDF) & Database Backups (.BAK)
Our Technical support people are split...
One wants to Add MORE memory and the other said "If you say this is the solution then I agree as you know more about SQL Server than I Do"
A look at the front of the server shows that the Disk access is 100%
mmmmmmmm
Viewing 15 posts - 1 through 15 (of 21 total)
You must be logged in to reply to this topic. Login to reply