How do folding VMs affect other VMs?

B

BassTek

Supreme [H]ardness
Joined
Jul 13, 2002
Messages
6,483
I'm nearly back to full power, with my three systems running and no tripped breakers in the computer room. All that remains is trying to figure out the VM setups for my two Q6600 systems.

One should be easy, I'll just run VMWare and have two Linux VMs to get the new A2 cores. The other system is tricky though, it's a Server 2008 host and it's main use is to run a test domain to help me study Server 2008. I'll have at least one server up all the time inside of Hyper-V, that being an IIS server. I'd like to add two Linux VMs but I'm worried it will cause resource conflicts with the other non-folding VMs. I'd like to run everything inside Hyper-V if possible since it plays nice with Server 2008 on both the host and guest side of things.

So after all that, my main question is: Has anyone run folding VMs alongside production VMs, and if so what were your experiences?

 
No idea if they would conflict but dont think so.
But one thing to try is to lower the priority of the folding VM's to minimum.
Then, so as long as rams not an issue due to short supply, they should give way to work VM's.

Luck ............ :D
 
I have 3 VM on my server, 2 for folding and 1 for Endian Firewall. So far, I never had problems but that's because the Endian Firewall VM is very light on resources.

It should work fine and if you notice slowdowns, just play with priorities like Tigerbiten said until the VM with your test domain run at a optimum speed.

 
I haven't really noticed any problems but the folding VMs are the most resource heavy VMs I run. I have one system running two Linux VMs for folding and a WinXP VM for a few other tasks. The XP VM doesn't need much in the way of resources but I've never had a problem with the other VMs causing it any trouble.

RAM is usually the most important part of running VMs. The more RAM you have, the better off they are going to be. The second most important thing about VMs is the priority they are set to run at. With VMWare in Windows, the VMs default to run at normal priority. This can cause problems as the VMs will take whatever cores and cycles are allocated to them and not give them up. The simple fix is to set the priority of the VMs to low. On my dedicated folding boxen I don't mess with the priorities at all. On my main box, I set the priorities to low otherwise trying to play games and such is practically impossible. Once the priorities are set to low I can't even tell the VMs are running. The VMs on low priority do take a small performance hit but it's more than worth it to have a box in which you can't even tell VMs are running in the background.

My two dedicated quads only have 2 gig of RAM and for now that's enough for 2 folding VMs and a third low priority VM as well as the host OS. These VMs are set to normal priority so there is a hit to the host VM. My main system has 4 gig of RAM and I can't even tell anything is running even with the GPU client running as well. I'd like to stick another 4 gig of RAM in the machine but that's something for later down the road and isn't actually "needed" right now.

 
Thanks for the replies. The box currently has 4GB of memory, it should be ok but I might have to bump it to 8GB. I currently have the IIS/SQL server taking 1GB, that leaves about 1GB per folding VM and 1GB for the host. I was hoping to squeeze a DC server in there as well but that might be pushing it. I also wanted to put a test exchange server in somewhere (gotta love Technet subscriptions) but that may have to be another box entirely.

Digging into the Hyper-V settings it looks like each VM has a 'relative weight' setting that I can use to prioritize them. There are also settings to reserve resources and set resource maximums but I'm thinking that could affect folding performance when none of the other VMs are doing anything.

 
You will not need 1 gig of RAM for the Folding VMs. The max I have allocated to a Folding VM is 768 meg and I scaled that back to 512 at a later time since the RAM was sitting there not being used. Remember, the only thing in the folding VM which needs RAM is the client itself and 512 meg is more than enough. You're not actively using anything in the VM except for folding so just about everything else could be paged to the swapfile and it wouldn't affect anything.

One last thing about VMs. If possible, it always helps to have them spread out over different physical hard drives if possible. I always try to keep my VMs on a different drive from my OS drive. This can make a big difference especially if you do keep the RAM allocation low for VMs and they do page.

 
Lowering the priority to low (or below normal) and sticking them to specific cores helps reduce contention. I'm doing this on my main rig with two VMs and the GPU client. I set the GPU client to real time and the VMs to low and assign them to specific set of cores (0&1 + 2&3 respectively). This works out perfectly for me.

 
You must log in or register to reply here.
Back
Top