INTERACT FORUM

Please login or register.

Login with username, password and session length
Advanced search  
Pages: [1]   Go Down

Author Topic: Best practices for running a Windows MC server in a virtual machine  (Read 1740 times)

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5181
  • "Linux Merit Badge" Recipient

So I'm setting up a windows virtual machine on my linux server solely for the purpose of running an MC for windows as a library server.  This is primarily because I need TV tuner support for recording and that's not in MC linux yet.  I've run linux virutal machines on windows hosts before, but I've never run a windows virutal machine for any length of time before.  I know several forum members (Inflatable Mouse, Glynor?) do run windows virtual machines in various contexts and I'm curious if anyone has any best practices to share in terms of setting up the VM environment, or for running MC in these kinds of contexts.  

I'm currently running the 64-bit version of windows 10 technical preview for evaluation purposes before buying another windows license, until I'm confident that I can get things working the way I need them to.  So I've gotten everything working with Virtual Box in a Debian Jessie host environment, but I'm curious if there are obvious general purpose optimizations I can make or any pitfalls I should be aware of.

An example of the kinds of optimizations I'm trying to make:

The machine is an i7 4790S (four actual cores, eight notional cores), and I've been fiddling with the number of cores passed through to the VM because I need it to do some transcoding, so I want to make sure it has enough grunt for the purpose.  Passing through three cores (out of eight) seems to benchmark significantly better than four cores with the same execution cap (I'm assuming because the virtual processors don't "count" the same way in a virtualization context and passing through four was choking the hypervisor?).  But two cores with no execution cap benchmarks almost as well as three cores with a 95% execution cap!  

I'm currently passing through 4G of RAM on the rationale that MC is more or less limited to 2G, but is that a correct assumption?  

If I want to reduce power consumption, do the power settings in the windows virtual machine actually do anything?  Or is it enough to handle the frequency scaling/power management on the Linux side?

I know this is one of those "how long is a piece of string" type threads (and might better be addressed on the virtualbox forums), but if anyone has any thoughts, suggestions, or gotchas on running MC in a windows virtual machine, I would greatly appreciate it.

Logged

mattkhan

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 3981
Re: Best practices for running a Windows MC server in a virtual machine
« Reply #1 on: April 03, 2015, 10:47:01 am »

I don't have anything to add to your question but just curious as to why you use virtualbox, I thought it was slower than xen/kvm, had that changed?
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5181
  • "Linux Merit Badge" Recipient
Re: Best practices for running a Windows MC server in a virtual machine
« Reply #2 on: April 03, 2015, 12:18:50 pm »

I don't have anything to add to your question but just curious as to why you use virtualbox, I thought it was slower than xen/kvm, had that changed?

I used virtualbox solely because I was familiar with it and it's hard to screw up.  I mentioned above that most of my prior virtualization experience has been windows host/linux guest and virtualbox works great in that context, so I used the tool I already knew.  I briefly looked into xen or kvm, but I couldn't find much discussion about modern windows os's in the docs (the official docs seem to peter out around windows 7).  

I'll try tinkering with them and see if I can improve performance.

At the moment, the JRMark of my virtualbox virtual machine (3400) is about the same as the JRmark on a four year old i7-2600k running win 7 bare metal (3500), so performance is currently relatively good, but better is always better.  The passmarks aren't quite as close (5234 for the Virtual Machine vice 7805 for the 2600K), but that's still pretty good for a VM. 

I'm most concerned about stability, but performance improvements are, of course, welcome.
Logged

mattkhan

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 3981
Re: Best practices for running a Windows MC server in a virtual machine
« Reply #3 on: April 03, 2015, 12:29:11 pm »

well fwiw I briefly ran windows 8.1 in a xen vm, it was pretty straightforward to setup once I had xen working. I gave up on it though because intel igp passthrough didn't work on xen 4.3, I think this is meant to be fixed in xen 4.5.
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5181
  • "Linux Merit Badge" Recipient
Re: Best practices for running a Windows MC server in a virtual machine
« Reply #4 on: April 03, 2015, 12:31:57 pm »

well fwiw I briefly ran windows 8.1 in a xen vm, it was pretty straightforward to setup once I had xen working. I gave up on it though because intel igp passthrough didn't work on xen 4.3, I think this is meant to be fixed in xen 4.5.

Ok, cool, glad for confirmation.  I'll give it a shot as I don't think I need access to the graphics processor in my server application, and I've got a little window of time.
Logged
Pages: [1]   Go Up