No. Putting the hypervisor on an SSD will not help it retrieve data from spinning disks more quickly. I think you’re thinking of a system that uses SSDs as a cache for spinners. You can’t get there like this. Also, that’s not what the BOSS card does. It has its best use in systems where one big RAID can’t solve the problem for some reason.

Use a RAID build with the highest number of spindles possible. Multiple smaller RAID systems is a micromanagement hell that wastes IOPS potential. Divide your singular array into two or more big partitions, if that makes you feel better. Monitor for drive failures. Keep good backups. Reallocation should not be a thing that you do. Expansion to handle new drives is OK, of course.

I guess I am not understanding the reason why this is a bad idea. Wouldn’t having the Host on SSD’s help to read the VM’s? Dell has the new BOSS controller that is designed exactly for this purpose as well as freeing up those two slots that would other wise be used for the host.

Also, I need some advice. I just got a new server and I was wondering what the best practice is for the disk space on the VM’s. Is it better to just create one large virtual disk and create the VM’s inside or put each VM on their own virtual? As an example, I have eight 4 TB drives. I have roughly 4 important systems: 3 SQL databases, and 1 File server each one using roughly 1.5 TB’s a piece, for now. Should I turn all 4 TB drives into one large 16 TB RAID 10 virtual disc and stick all the VM’s in there or should I split it into 4 sets of 4 TB virtual disks each in RAID 10 and stick each server on their own virtual disc? I can see that the advantage of putting it in the large 16TB would be that if any of them grow past the 4 TB then I can just reallocate accordingly, it is more dynamic, however there seems to be a higher possibility of losing all the VMs if enough drives go down at once or if I need to expand more than one machine, I have to backup and restore ALL VMs. On the other hand putting them each in their own partition makes it so if for what ever reason need to expand a VM past that 4 TB, I can just back up that one VM, then replace just those 2 physical disks with larger drives and throw the VM back on it. I can do this without even affecting the other VMs. I am just not sure how this works on performance of the RAID controller and the Read and write abilities.

Your Thoughts?

Most intelligent comment I’ve read on this particular topic in like…………………….EVER !

Every single one of my Windows Server VMs has a C: VHDX under 40GB, with the exception of the SQL VMs because SQL. Most of my Server Core VMs are under 15GB. If routine patches and upgrades are causing your VM to balloon over 40GB, then something is broken or you’re putting more in the C: partition than just Windows.
I stand by my statement.

That doesn’t make any sense. If you have a single Hyper-V host running its own domain controller, then the only way to permanently lock yourself out is by intentionally disabling cached credentials AND intentionally disabling the local host administrator account AND if the virtual domain controller fails or isn’t set to auto-start. Or, in other words, if you intentionally architect for failure, you’re at a higher risk of failure. Even if you have done all of the above, you should still be able to restore your virtualized domain controller from backup to another machine, even a Windows 10 desktop, and that will be enough to get everything going again. It’s just a waste of money to keep a physical DC as a security blanket against a problem that literally no one in a small environment ever has any reason to cause for themselves.

I should have added that your VM’s will boot up without having to logon to the Hyper-V host, so it technically will work and not lock you out assuming you have your VM’s setup to power up on server boot. I would recommend though that you definitely leave yourself a break glass local account in case the VM domain controller does not boot up for whatever reason and then you cannot logon to the Hyper-V host to troubleshoot it. For this reason I prefer to keep at least one physical GC on the network. But you can do it either way, just make sure you have that local admin account for safety.