A best practices question i guess.
Say I have 10 new blades every one with 2x16 Cores and 256GB ram and FC storage to the tune of 10x2TB luns. Thin on back end. Networking is strictly external.
So in total I would have: 320 CPU's , 2560GB of ram and 20TB disk.
They are all configured into a computing cluster with DRS and HA.
Now I have 4 different groups that want a big bite of my apple and 2 that want a small nibble.
Should the allocation model be the same for all of them? I.e. Pay as you go with the exception of the disk allocated strictly to the different groups by using storage profiles and dividing along luns.
Every org should have their own VDC?
Is there a better way to achieve more flexible results?
This really comes down to what you have to guarantee. If you have to make sure they all get a very specific minium, then you might want to make one cluster and create different reservation values for each need.
However if you can get along with general VM quotes, and basic CPU and mem reservation numbers - then Pay Go is the easy way. (And what I try to use whenever possible; as it allows you to get the most out of the equipment you have).
What I am really trying to figure out ahead of time are the possible pitfalls of having multiple provider vcd's exist on same hardware cluster. To what level do they co-exist and not "trample" on each other. or is there a more elegant way to do this that might be more scalable.