Just after a bit of quick advice...
I've got a Dell R610 with 2 x 146Gb disks and 4 x 600Gb disks.
My inclination was to set up 1 x RAID1 on the 2 x 146Gb and 1 x RAID5 on the 4 x 600Gb disks.
My colleague said it might be better to set up 3 x RAID1 with the 3 pairs of disks?
Also, is there any benefit these days with splitting the volumes i.e. virtual guest OS / virtual guest page file / splitting the virtual disks onto seperate RAID LUNS on the virtual host?
Hope someone can shed some light...
administrativly no. its a pain to split them.
if you had disks that might be fast, say ssd, then you might use that for your vswap etc.
I thought it would be a pain to split them to be honest and I can't say I've read anywhere from VMware best practices that recommend's it....
you may be able to get a minor improvement in performance with an individual configuration. However, this really depends on many factors and may not be worth the effort in most cases. I'd go with the RAID configuration you mentioned (RAID1 + RAID5) and don't bother about splitting the VM's virtual disks. The OS disk does not require much performance anyway.
What you should make sure is that you have a RAID controller with the BBU (battery-buffered write cache) option. This makes a huge difference in disk performance.
Thanks for this great useful information.
I'll definately check up on the BBU - the RAID controller I have is a Dell PERC H700 Integrated controller....
I was under the impression that splitting disks like in the old days was now a thing of the past with the hypervisor taking care of requests etc...
For a local array and depending how what you OS workload is splitting the drives will get you a very minor performance improvement.
they way id go if you could would be this
get a 1 GB SD load ESXi on it.
get 2 more 600 GB drive and either
6 X RAID 5 -- medium performance lots of space around 2793.85 TB
6 X RAID 0+1 -- Stripped mirrors safety and performance at the cost of sapce 1676.38 TB
We run all our R610 with SD cards and either local or Fiber disks this way.