Patch: cpu utilization monitor.

[linas at austin.ibm.com]

On Wed, Mar 17, 2004 at 09:36:23AM -0800, Greg KH wrote:
>
> On Wed, Mar 17, 2004 at 09:32:16AM -0800, Dave Hansen wrote:
> > On Wed, 2004-03-17 at 09:27, Greg KH wrote:
> > > On Tue, Mar 16, 2004 at 05:38:06PM -0600, ahuja at austin.ibm.com wrote:
> > > > This patch adds the framework required by performace team and on demand
> > > > computing. At this point only the important bits/framework are covered.
> > >
> > > Is this what the ckrm people need?  If not, that is what the kernel is
> > > going to be supporting soon, so I would suggest you look into their
> > > solution.
> >
> > What are the CKRM people doing?  Scheduling classes based on usage of
> > certain performance counters?
>
> They are doing "resource management" type stuff for almost everything in
> the kernel, including scheduling and performance monitoring.  See their
> documentation for more details, but it sounds like it is exactly what
> this author is looking for.

This patch differs from other efforts in that it gets data directly from
the hypervisor.  Think multiple virtual cpus running on one physical cpu.
The traditional tools, whether CKRM or top or vmstat, are blind to the
fact that any given 'virtual cpu' might be getting only 10% of the physical
cycles in one hypervisor time-slice, and 90% in another.

Very crudely, its sort-of like VM on the 390/zSeries.  Your kernel may
think its 100% busy, but in fact it might be getting only 1% of the actual
physical hardware cycles.  The goal here is to be able to report the
fraction of the total physical cycles, and do so on a HZ or even sub-HZ
level of granularity.

--linas


** Sent via the linuxppc64-dev mail list. See http://lists.linuxppc.org/