Hi,
I'd appreciate a sanity check on a proposed deployment please 
SAN - Dell MD3220i
- 13 x 10K 600 GB SAS drives
- 3 x 149 GB SSD drives
- 8 x spare bays
The plan is to deploy the SSD drives in RAID 5 giving up to 300 GB for databases, highly transactional VMs, etc. This can be the high performance "fix" for those VMs that need it.
I want to deploy one global hot-spare leaving me with 12x 10K 600 GB drives and in the future, deploy further 10K 600 GB SAS drives in the empty bays or possibly some larger capacity NL-SATA drives.
My first thought is a single disk group of 12 drives in RAID 6 giving me up to 6 TB. I would then create three LUNs of 2 TB each, each with a fully-sized VMFS volume. This should give me plenty of IOPS because of the number of spindles.
I'm hoping RAID 6 is the right choice because the rebuild time shouldn't be too bad with 10K 600 GB SAS drives and it avoids the performance degradation of the RAID 5 rebuild process.
If I was to split the available drives into smaller disk groups of RAID 6 I'd lose a lot of capacity.
For the future, I'd probably like to create another RAID 6 disk group of eight disks.
Does all this sound sensible?
Thanks for looking
Further to the above, I've decided to implement a single disk group in RAID 6 over all 12 disks with one more drive as a hot spare - still very interested to hear anyone's opinion though.
I've created 6 virtual disks of 1 TB each (the last one is around 500 GB) - I've based this figure on the idea that I'll have 20 VMs per LUN each at around 40 GB in space each on average and giving me 20% free space for snapshots, extending disks, etc.
I specified a segment size of 128K for each VM though I couldn't find any "best practice" approach for doing this so left it up to the SAN management software to decide
It would be nice to see some sort of best practice guide from Dell and/or VMware rather than having to use Google and hope for the best.
Hi,
Thanks for your reply
I only had budget for the three SSDs but RAID 10 wouldn't give any appreciable performance gains compared with what the SSDs will give anyway - I'm using the Dell SLC enterprise ones so I'm expecting some good results! :smileycool: I don't think the 10K drive can be a hot-spare for the SSDs but I would hope the SSDs don't really need one especially as I'm using RAID 5.
If I trusted RAID 5 for mechanical disk drives I would have preferred two such disk groups as you suggest but availability and reliability are extremely important - performance much less so. As a result, I think losing an extra drive to give yourself protection for a second drive failure is no problem at all and hopefully, if I do need plenty of IOPS I have the SSD as an option or the number of spindles on the SAS disk group will make up for not going RAID 5. Perhaps Seagate will release some 15K 2.5" drives before I fill out the remaining 8 bays as well!
Experience-wise, I've been running a SAN with two disk groups, a RAID 6 and a RAID 5 - 15K SAS 300GB drives. There was a drive failure (on the RAID 6 ironically) and by the time I'd verified it, the hot spare had kicked in and the array had rebuilt. Replacing the drive, by the time I returned home after visting the data centre to replace the drive (after ordering a new one for next-day delivery), the array had just about finished rebuilding! The point is, even though a RAID 6 drive had failed, it was still protected and would be even during the rebuild process. If a RAID 5 drive had failed, then all that time taken to order a new drive and knowing it was rebuilding, and knowing that this was the most stressful experience for the array, is something I would always try and avoid in the future. Backups are fine of course but restoring that much data and getting it all up and running again is something I wouldn't choose to do if I could avoid it in any way!