Re: [libvirt] [PATCH 06/10] vcpubandwidth: introduce two new libvirt APIs

[Taku Izumi <izumi taku jp fujitsu com>]

On Wed, 06 Jul 2011 10:40:47 +0530
"Nikunj A. Dadhania" <nikunj linux vnet ibm com> wrote:

> On Wed, 6 Jul 2011 10:18:05 +0900, Taku Izumi <izumi taku jp fujitsu com> wrote:
> > 
> > > Nikunj is correct here.  We only need to ensure that the bandwidth is
> > > distributed equally between all of the cpus.  This can be accomplished
> > > internally by creating a cgroup for each vcpu and ensuring that
> > > 'cpu.share' is the same for each vcpu cgroup.
> > 
> > 
> > When taken together, virsh I/F and domain XML schema 
> > are the following:
> > 
> > #virsh schedinfo VM --set cpuperiod=100000 --config
> > #virsh schedinfo VM --set cpuquota=50000 --config
> > 
> > ..
> > <vcpu>4</vcpu>
> > <cputune>
> >   <period>100000</period>
> >   <quota>50000</quota>
> > </cputune>
> > 
> > When specified above, the qemu driver creates cgroup for
> > each vcpus and specified as follows:
> > 
> > The quota value is divided by the number of vcpus.
> >  vcpu#: (quota/period) 
> >  vcpu0: 12500/100000 
> >  vcpu1: 12500/100000
> >  vcpu2: 12500/100000
> >  vcpu3: 12500/100000
> >
> I am with you till this point. This is what I was visualizing the
> interface as, thanks for putting it.
>    
> > But, these values cannot be confirmed form the domain 
> > XML file.
> Do not understand what do you mean here.

 Sorry. What I wanted to say is:

 In case of the above example, the administrator cannot know that
 each vcpu's quota is specified to 12500, because domain's XML
 definition only shows that domain's quota is 50000.
 Additionally, this information cannot be gotton via virsh command
 because it is only displayed as the following:

 # virsh schedinfo VM
 Scheduler      : posix
 cpuperiod      : 100000
 cpuquota       : 50000

 It was a little anxious that actual setting is encapsulated.

 Anyway, I understand what you intend. Thank you.

-- 
Taku Izumi <izumi taku jp fujitsu com>