> Hi,
>
> in my application running on a MSVS 2005 I need to calculate max CPU
> allocation that each of VMs is allowed to consume.
>
> If I correctly understand how VS works I need to get the value of the
> "Maximum Capacity (% of one CPU)".
>
> Unfortunately I can't find API to get this value. Both:
> 1. MaximumSystemCapacity Property of the IVMAccountant Interface
> and
> 2. settings/host/resource_control/cpu/cap in VM's vmc file
> have the same value which seems equal to "Maximum Capacity (% of system)"
>
> Does anybody know how to get the former or calculate it from the other?
>
> Here's the problem, because:
> A. on a dual-core AMD machine they are the same (factor of 1 for one
> physical processor?). It doesn't make sense - the VM is able to consume 1
> core at most if you set the limit to 100% of a CPU, yet VS tells you it
> utilizes 100% of system resources (which are two cores) while the other core
> is idle.
>
> B. on a dual CPU Xeon machine, % of system is half % of CPU (factor of 2 for
> two physical processors?)
>
> The conclusion (divide by number of physical processors) seems obvious but
> I'm reluctant to it because in both cases the server has 2 processors to run
> VMs on (why would it matter if the cores lie on the same physical socket or
> two?).
>
> Could someone please enlighten me on that?
>
> I'm using MS VS 2005 R2 version "1.1.465.292 EE R2"
>
> Thanks in advance for any info.
>
> Regards,
> Filip

>
> Multicore systems work the same as multi-cpu systems, the system
> resources are calculated for all cores, on a dual-core, 1 VM can get
> 100% of one core == 50% of system resources, for a quad-core, 1 VM
> get's 100% of 1 CPU == 25% of the system resources.
>
> --
> Cheers,
> Steve Jain, Virtual Machine MVP
> http://vpc.essjae.com/
> I do not work for Microsoft.
>

>and not per core.
>"Filip Barczyk" wrote:
>> Hi,
>>
>> in my application running on a MSVS 2005 I need to calculate max CPU
>> allocation that each of VMs is allowed to consume.
>>
>> If I correctly understand how VS works I need to get the value of the
>> "Maximum Capacity (% of one CPU)".
>>
>> Unfortunately I can't find API to get this value. Both:
>> 1. MaximumSystemCapacity Property of the IVMAccountant Interface
>> and
>> 2. settings/host/resource_control/cpu/cap in VM's vmc file
>> have the same value which seems equal to "Maximum Capacity (% of system)"
>>
>> Does anybody know how to get the former or calculate it from the other?
>>
>> Here's the problem, because:
>> A. on a dual-core AMD machine they are the same (factor of 1 for one
>> physical processor?). It doesn't make sense - the VM is able to consume 1
>> core at most if you set the limit to 100% of a CPU, yet VS tells you it
>> utilizes 100% of system resources (which are two cores) while the other core
>> is idle.
>>
>> B. on a dual CPU Xeon machine, % of system is half % of CPU (factor of 2 for
>> two physical processors?)
>>
>> The conclusion (divide by number of physical processors) seems obvious but
>> I'm reluctant to it because in both cases the server has 2 processors to run
>> VMs on (why would it matter if the cores lie on the same physical socket or
>> two?).
>>
>> Could someone please enlighten me on that?
>>
>> I'm using MS VS 2005 R2 version "1.1.465.292 EE R2"
>>
>> Thanks in advance for any info.
>>
>> Regards,
>> Filip
>
>Greetings Filip,
>
>I'm not sure if this answers your questions or serves as info, however it
>might :)
>
>As far as I am aware Virtual Server 2005 works only per physical processor,
>and not per core. However your host system will operate with cores, meaning
>that your host will allocate any host associcated task on the core that is
>least busy.
>When that is said, Virtual Server 2005 will not gain any benefit from
>running on Hyperthreading CPU's. - as these will work, from virtual server,
>as different cores.
>
>I hope that gave some input on atleast what to search for..
>
>Kind Regards
>Jesper

>Hi Steve,
>
>That is exactly my problem, on my single Opteron DualCore machine, I gave a
>VM 100 % of CPU and it shows 100 % of system, not 50% as you'd expect.
>
>Do you have any ideas why it is so? Maybe some patch for multicore support
>is needed?
>