Windows guest CPU socket/core recognition

Windows guest CPU socket/core recognition

[Zdenek Kaspar]

Hello everyone,

I guess I'm not the first one who hit the problem with Microsoft's
licensing model..

Nowadays the common single or dual quad-core workstation can't be fully
used because it's limited by example: license up to 2 physical
processors. Such VM acts like 4-way or 8-way machine.

Is there any way howto expose CPUs differently for this kind of problem?

TIA, Z.

Re: Windows guest CPU socket/core recognition

[Brian Jackson]

On Monday 17 August 2009 22:28:35 Zdenek Kaspar wrote:
> Hello everyone,
> 
> I guess I'm not the first one who hit the problem with Microsoft's
> licensing model..
> 
> Nowadays the common single or dual quad-core workstation can't be fully
> used because it's limited by example: license up to 2 physical
> processors. Such VM acts like 4-way or 8-way machine.


Nine times out of ten, a single cpu guest is going to be a better option than 
a smp/multie core guest. I've seen idle windows guests go from using nearly 
200% cpu for -smp 2 to ~5-10% for -smp 1. Unless your guest is actually using 
all that cpu all the time, you're going to be wasting a decent amount of 
cycles.



> 
> Is there any way howto expose CPUs differently for this kind of problem?


There have been patches (from Andre Pryzwara and maybe others) to support 
multi-core vs mult-socket smp.


> 
> TIA, Z.
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: Windows guest CPU socket/core recognition

[Zdenek Kaspar]

Brian Jackson napsal(a):
> On Monday 17 August 2009 22:28:35 Zdenek Kaspar wrote:
>> Hello everyone,
>>
>> I guess I'm not the first one who hit the problem with Microsoft's
>> licensing model..
>>
>> Nowadays the common single or dual quad-core workstation can't be fully
>> used because it's limited by example: license up to 2 physical
>> processors. Such VM acts like 4-way or 8-way machine.
> 
> 
> Nine times out of ten, a single cpu guest is going to be a better option than 
> a smp/multie core guest. I've seen idle windows guests go from using nearly 
> 200% cpu for -smp 2 to ~5-10% for -smp 1. Unless your guest is actually using 
> all that cpu all the time, you're going to be wasting a decent amount of 
> cycles.

Yes, without proper use it's waste. My guest is 64bit and represents
node in computing cluster working on assigned jobs (not MPI etc..)

>> Is there any way howto expose CPUs differently for this kind of problem?
> 
> 
> There have been patches (from Andre Pryzwara and maybe others) to support 
> multi-core vs mult-socket smp.

I will check these patches, thank you (and Dor Laor) for the hint!

Re: Windows guest CPU socket/core recognition

[Dor Laor]

On 08/18/2009 06:28 AM, Zdenek Kaspar wrote:
> Hello everyone,
>
> I guess I'm not the first one who hit the problem with Microsoft's
> licensing model..
>
> Nowadays the common single or dual quad-core workstation can't be fully
> used because it's limited by example: license up to 2 physical
> processors. Such VM acts like 4-way or 8-way machine.
>
> Is there any way howto expose CPUs differently for this kind of problem?

It was discussed on 
http://www.mail-archive.com/kvm@vger.kernel.org/msg18490.html but not 
yet merged.


>
> TIA, Z.
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Windows guest CPU socket/core recognition

[Andre Przywara]

Dor Laor wrote:
> On 08/18/2009 06:28 AM, Zdenek Kaspar wrote:
>> Hello everyone,
>>
>> I guess I'm not the first one who hit the problem with Microsoft's
>> licensing model..
>>
>> Nowadays the common single or dual quad-core workstation can't be fully
>> used because it's limited by example: license up to 2 physical
>> processors. Such VM acts like 4-way or 8-way machine.
>>
>> Is there any way howto expose CPUs differently for this kind of problem?
> 
> It was discussed on 
> http://www.mail-archive.com/kvm@vger.kernel.org/msg18490.html but not 
> yet merged.

I have a reworked version almost ready. The AMD part is working fine, 
but I still have troubles with the Intel part and threads vs. cores (in 
the guest).
I hope I can fix this still this week.

Regards,
Andre.

-- 
Andre Przywara
AMD-OSRC (Dresden)
Tel: x29712