FoldOppo, What Would You Recommend?

"TheRealBicycleBuck" (therealbicyclebuck)
04/15/2020 at 08:51 • Filed to: None

I’ve mentioned that I have a small, old server available to me. I went through yesterday and shut down all of the VMs not being used, so now I have some space to work with. The server is a dual Xeon X5650, 2.66Ghz, 12M Cache,Turbo, HT, 1333MHz with 64 GB RAM. It’s running VMWare, so I can spin up a VM or two (or more) and set them to folding.

I’m looking to optimize the VM configuration for the folding task. Would you give one VM a lot of resources or would you set up a cluster of VMs and distribute the tasks? What OS would you run and how would you optimize the configuration?

A network of cars for looking...

DISCUSSION (9)

themanwithsauce - has as many vehicles as job titles > TheRealBicycleBuck
04/15/2020 at 09:37

Since CPU units tend to be smaller, I think more VMs would be better than one VM with all the resources.

TheRealBicycleBuck > themanwithsauce - has as many vehicles as job titles
04/15/2020 at 10:36

Ok. that’s food for thought. Any opinions on the OS or how mu ch memory to allocate?

themanwithsauce - has as many vehicles as job titles > TheRealBicycleBuck
04/15/2020 at 11:02

Arguably linux would be best for lowest overhead. I don’t think this takes much ram at all. Don’t know what your server has but no more than a gig or two per VM should be necessary. Cooling and power capabilities will probably be your li mi ting factors.

TheRealBicycleBuck > themanwithsauce - has as many vehicles as job titles
04/15/2020 at 11:23

The server has 64gb of RAM, so it has lots of room. I’m wondering where the point of diminishing returns lies. Let’s say I dedicate 32gb to folding. Would 16 VMs running linux with 2gb RAM each be the best configuration? Or would 8 VMs with twice the RAM be better? At what point does the processor become the limiting factor?

Future Heap Owner > TheRealBicycleBuck
04/15/2020 at 11:49

Each VM has some amount of overhead (e.g. the kernel has to run and that’s not doing any folding itself), so I’d think one VM with as much CPU as you can spare would be the most efficient.

Manwich - now Keto-Friendly > TheRealBicycleBuck
04/15/2020 at 11:54

I would say skip the VM and run it directly on the server... either with windows 10 or with Xubuntu

And if you want to make some decent numbers, stick the best graphics card you have into it.

VMs make sense for servers that are lightly used most of the time. For running Folding@home, it’s running full out all the time. In cases like that, a VM just adds overhead and no real net benefit.

TheRealBicycleBuck > Manwich - now Keto-Friendly
04/15/2020 at 12:07

Thanks for the suggestions, but t he server was built to run VMs and is currently hosting one that I need to keep online. Wiping it to do a fresh install of Win10 isn’t going to happen.

It’s also a rack-mounted server, not a tower. I’m not sure there’s space in there for a dedicated graphics card.

404 - User No Longer Available > themanwithsauce - has as many vehicles as job titles
04/15/2020 at 12:14

Or a large VM and Dockerize it?

https://hub.docker.com/r/linuxserver/foldingathome

Manwich - now Keto-Friendly > TheRealBicycleBuck
04/15/2020 at 12:59

Ah... okay. Well in that case, then I’d say assign the minimum resources to that VM you have to keep going and just have all the remaining CPU resources assigned to a 2nd VM, along with ‘just enough’ disk space and memory.

Folding@Home is a CPU/GPU (if present) hog. It doesn’t need all that much RAM or disk space.

And for the OS in that VM, run something with low overheads like Lubuntu or Xubuntu. I personally use Xubuntu on a couple of older laptops and I like the interface better than Ubuntu, which I have on my main personal home desktop.

And given it’s a rack mount server, yeah, it’s unlikely you’ll fit anything except a low-profile, low-power graphics card in it. So don’t bother. I thought your server might be one with a tower case.

If you have an old workstation with a tower case, that would likely be a better candidate for Folding@Home from a performance per watt perspective IF you put in at least a half-decent graphics card in it.

If you want to rack up the points, having a system with one or more powerful graphics cards is the way to do it.... provided you have the slots and provided the power supply can handle it.