We currently have a MD 3220 with 24*600GB HDDs, with 2 Poweredge 620s attached. They run a clustered Hyper-V. This has got some irresolvable software issues, so we are buying more kit (needed anyway) and creating a new cluster. We are getting 2 more 620s, but more to the point an MD1200 with 12*3TB drives, and a 1220 with 24*900GB drives.
So, we are going to end up with 4 620s, 24*600GB, 24*900GB, 12*3TBs. I want to get the disk configuration right, as I am in no way convinced we got it correct the first time round - I simply made a single RAID 10 array with 20 disks in (4 Hot-swaps). We support all sorts of VMs - Exchange, SQL , File, Remote Desktop, DCs etc. Probably running about 50 VMs currently.
Are there best practices as to number of virtual disks? Am I best for example, to put 4 disks in a single 1.2TB Raid 10, or should I stick with small amount of large disks (EG 10 disks in each raid 10)?
I've never found anything helpful from Dell about Cache size - there is a document which discusses Cache Size vs usage (Database, file store), but these seem quite old and do not mention Hyper-V - is it sensible to have a group of disks for SQL, another group for Exchange, another for generic files and adjust Cache Size on each group of disks - or is there a generic Hyper-V recommended setting?
While I'm here, is there a significant boost for the high-performance mode?
Seeing as you have different needs for different virtual disk I would look at having a few different disk groups. With the different disk groups you can still have they all owned by the cluster so that all host will have access to all virtual disk. Now with having multiple disk groups you will need to make sure that you account for data growth so that you can assign the correct amount of drives so that you don’t run out of space to any one disk group. In most cases you can always add more drives from your free space to any disk group. The last thing to remember is that you don’t want to have any virtual disks span multiple enclosures. I have seen where customers have done this & it has had a real performance impact on the data.
Please let us know if you have any other questions.
I too have been looking for best practices and setup suggestions with the MD3220. I have a brand new one with 24x146Gb SAS 15k drives. My MD3220 is SAS connected using dual SAS cards in each server for redundancy. My VMs will be two SQL instances and 3 application servers with about 20 connected users. Any my disk size requirements are pretty low as well at about 250Gb so my MD3220 has way, way more storage than I need even with growth. I called Dell support to get their suggestion and the guy actually said "well you can do it any way you want to" when I asked how I should configure with my existing VM needs. I could have gotten the same damn answer from my wife.
Thanks for any input as for making a single/large RAID10 volume or multiple disk groups.