SQL Virtualization - SAN setup

hi all

we will soon be virtualizing our SQL server using vmware and moving it to a SAN.

what i am worried about is configuring the virtualisation wrong so that the SAN disc subsystem is not utilised properly.

what i will be doing is breaking the 12 physical discs into 4 logical drives

c: 2 discs -raid 1 - os
D: 6 discs - raid 5 - databases
E  2 discs - raid 1 - logs
F  2 discs - raid 1 - Tempdb

regarding disc's D,E and F, do i simply map these to drive C. as i assume if these discs are setup as virtual drives  the virtual machines disks are simply files stored on the file system and i will lose the benefit of  the Raid setup?

can someone advise? all advice appreciated!
Who is Participating?
I wear a lot of hats...

"The solutions and answers provided on Experts Exchange have been extremely helpful to me over the last few years. I wear a lot of hats - Developer, Database Administrator, Help Desk, etc., so I know a lot of things but not a lot about one thing. Experts Exchange gives me answers from people who do know a lot about one thing, in a easy to use platform." -Todd S.

Alex GalbraithSolutions ArchitectCommented:
You have a couple of options here, your choice will be partly dependant on if you are sharing the SAN with other VMs:

1. From your SAN present several LUNS, formatted as VMFS:
LUN1 - RAID1 (VMFS Formatted)
LUN2 - RAID1 (VMFS Formatted)
LUN3 - RAID5 (VMFS Formatted)
Then configure 4 VMDKs for this VM as follows:
C:\ VMDK on LUN1
D:\ VMDK on LUN3
E:\ VMDK on LUN2
F:\ VMDK on LUN2
This will provide you with the performance and redundancy you need, with the flexibility of being fully virtualised.

2. From your SAN present several LUNS, one is VMFS for the OS and the rest as RDMs (raw data mapping):
LUN1 - RAID1 (VMFS Formatted)
Then configure your VM as follows:
C:\ VMDK on LUN1
D:\ RDM to LUN4
E:\ RDM to LUN2
F:\  RDM to LUN3
This will provide maximum performance from having direct access to the LUNs, but you will lose all of the benefits of virtualised disks, including the ability to VMotion, though it would allow you to later cluster the VM to a physical box if you so desired.

Personally I would go with option 1 as you will have the best balance of performance, redundancy and flexibilty. Feel free to split out the TempDB onto a seperate LUN from the logs if you like (even better performance again), or if you have many VMs on your storage / are limited on cost and LUNs, then use one VMFS LUN for OS, Logs and TempDB (less redundancy and performance).

Experts Exchange Solution brought to you by

Your issues matter to us.

Facing a tech roadblock? Get the help and guidance you need from experienced professionals who care. Ask your question anytime, anywhere, with no hassle.

Start your 7-day free trial
malraffAuthor Commented:
hi jkagalbraith

thanks for your input,
there wil be no other machine on the san, so if i go with option A - would i lose much performance?

and considering i have 12 * 300k 15rpm discs, do you think i am spliting them wisely with best raid selection for each?

i was tempted to go raid 10 for the databases, but decieded not mainly due to the fact iv never configured a system as such and read that some raid controller really do not like it?
"malraff", what you figured out to configure is perfect really. Configure those "drives" on your SAN as LUNs, then use 'Raw Device Mapping' to 'present those LUNs to your VM, as "jkagalbraith" mentions above. Once presented, they will show up in disk mgmt on your VM and you add the disks/format like usual. I agree with you in staying away from a RAID10; besides potential controller issues, you lose 2 disks worth of storage (intead of only 1 with RAID5). Your logs/tempdb volumes have no need to be in RAID5, solely RAID1 for disk failure recovery purposes. You most certainly have a pretty good rollout plan.

Best of luck!
Determine the Perfect Price for Your IT Services

Do you wonder if your IT business is truly profitable or if you should raise your prices? Learn how to calculate your overhead burden with our free interactive tool and use it to determine the right price for your IT services. Download your free eBook now!

malraffAuthor Commented:
cheers coolsport, good to hear someone thinks im on the right track !
Alex GalbraithSolutions ArchitectCommented:
I have just replied to someone regarding choice of RAID 5/10 here:


RAID10 will give better write performance at a greater overall cost.

If you have no intention to be able to use VMotion then the RDMs will give you better performance, however, unless your server is being completely hammered, the performance hit of full VMFS virtualisation will be minimal to the point where you will not likely tell the difference.
You can use VMotion with RDM...just VMotioned 3 servers that have RDMs associated with them.

Alex GalbraithSolutions ArchitectCommented:
It does work, but its flakey at best and wont work if you have clustered VMs (not technically supported anyway).
Hmm..It's worked for me, unhindered, for the past 2yrs; agree about the cluster, though.

malraffAuthor Commented:
jkagalbraith: and regarding info iv read with some controller issues with raid 10 is that mostly uncommon now?
also does raid 10 only workin multiples of 4? eg i have 6 discs for the databases, but i dont see how raid 10 would utilise all    
and would a 6 discs raid 5 be similar in speed to 4 disc raid 10? or is that a pointless comparison?
Yes, you won't have any problem creating a RAID10 with 6 disks. It will be configured as a mirror of 2 sets of 3 disk in each set. For performance comparison between 6 disk RAID 10 and 4 disk RAID 5, i would say you won't see too much difference, but the more underlying disks you have, the faster the write/read operation generally will be.
Alex GalbraithSolutions ArchitectCommented:
for 6 (RAID10) v 4 (RAID5) you will see a bit faster writes on RAID10, and obviously the RAID10 will provide better redundancy (minimum 2 disks can be lost and no data loss).
RAID10 works with 2xN drives
Some HBA has RAID1E that allows N drives in a mirroring capability
Regarding RAID 5...please don't do that for your DATA...The write penalty is such a pain for random write usage that you should keep in mind that RAID 5 is only for Backup/Archive/WORM usages
A 6 drives RAID5 would perform better than a 4 drives RAID 10 on all read operations if the RAID5 partition is aligned (otherwise : every io may involve 2 drives to achieve)...but random write operations would be 10x SLOWER than their RAID10 equivalent (even with a HBA write cache backed by a battery)

I would have only 2 LUN:
  4 in RAID10 OS / Swap / Apps / TempDb / Logs / Backup2Disc for a 600GB using a 1MB stripe size
  8 in RAID10 Index / Data for a 1.2TB db using a Max(64KB; 8x(db page size)) maximum for stripe size
Nota: Always align your partition to a stripe size multiple (1MB minimum) using Diskpart.exe (Win2003 aligns to 31.5KB by default)
malraffAuthor Commented:

alot of info to digest, BigSchmuh> i do have a gut feeling raid 10 may be the way to go, i do have a week to test so i guess i can try a few configs first!

It's more than this solution.Get answers and train to solve all your tech problems - anytime, anywhere.Try it for free Edge Out The Competitionfor your dream job with proven skills and certifications.Get started today Stand Outas the employee with proven skills.Start learning today for free Move Your Career Forwardwith certification training in the latest technologies.Start your trial today
Microsoft SQL Server 2005

From novice to tech pro — start learning today.