Re: [ckrm-tech] [RFC][PATCH] UBC: user resource beancounters

[Kirill Korotaev <dev@sw.ru>]

Chandra Seetharaman wrote:
> On Thu, 2006-08-17 at 17:55 +0400, Kirill Korotaev wrote:
>
> > > On Wed, Aug 16, 2006 at 07:24:03PM +0400, Kirill Korotaev wrote:
> > >
> > >
> > > > As the first step we want to propose for discussion
> > > > the most complicated parts of resource management:
> > > > kernel memory and virtual memory.
> > >
> > > Do you have any plans to post a CPU controller? Is that tied to UBC
> > > interface as well?
> >
> > Not everything at once :) To tell the truth I think CPU controller
> > is even more complicated than user memory accounting/limiting.
> >
> > No, fair CPU scheduler is not tied to UBC in any regard.
>
>
> Not having the CPU controller on UBC doesn't sound good for the
> infrastructure. IMHO, the infrastructure (for resource management) we
> are going to have should be able to support different resource
> controllers, without each controllers needing to have their own
> infrastructure/interface etc.,
1. nothing prevents fair cpu scheduler from using UBC infrastructure.
  but currently we didn't start discussing it.

2. as was discussed with a number of people on summit we agreed that
  it maybe more flexible to not merge all resource types into one set.
  CPU scheduler is usefull by itself w/o memory management.
  the same for disk I/O bandwidht which is controlled in CFQ by
  a separate system call.

  it is also more logical to have them separate since they
  operate in different terms. For example, for CPU it is
  shares which are relative units, while for memory it is
  absolute units in bytes.

> > As we discussed before, it is valuable to have an ability to limit
> > different resources separately (CPU, disk I/O, memory, etc.).
>
> Having ability to limit/control different resources separately not
> necessarily mean we should have different infrastructure for each.
I'm not advocating to have a different infrastructure.
It is not the topic I raise with this patch set.

> > For example, it can be possible to place some mission critical
> > kernel threads (like kjournald) in a separate contanier.
> I don't understand the comment above (in this context).
If you have a single container controlling all the resources, then
placing kjournald into CPU container would require setting
it's memory limits etc. And kjournald will start to be accounted separately,
while my intention is kjournald to be accounted as the host system.
I only want to _guarentee_ some CPU to it.

Thanks,
Kirill
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/