Re: Summary of resource management discussion

From: "Paul Menage" <menage@google.com>
To: vatsa@in.ibm.com
Cc: xemul@sw.ru, dev@sw.ru, pj@sgi.com, sam@vilain.net,
	ebiederm@xmission.com, winget@google.com, serue@us.ibm.com,
	akpm@linux-foundation.org, linux-kernel@vger.kernel.org,
	ckrm-tech@lists.sourceforge.net, containers@lists.osdl.org
Subject: Re: Summary of resource management discussion
Date: Thu, 15 Mar 2007 04:24:37 -0700	[thread overview]
Message-ID: <6599ad830703150424t3478cd55mf9d2699f3669c9f0@mail.gmail.com> (raw)
In-Reply-To: <20070312124226.GD17151@in.ibm.com>

On 3/12/07, Srivatsa Vaddagiri <vatsa@in.ibm.com> wrote:
>         - (subjective!) If there is a existing grouping mechanism already (say
>           tsk->nsproxy[->pid_ns]) over which res control needs to be applied,
>           then the new grouping mechanism can be considered redundant (it can
>           eat up unnecessary space in task_struct)

If there really was a grouping that was always guaranteed to match the
way you wanted to group tasks for e.g. resource control, then yes, it
would be great to use it. But I don't see an obvious candidate. The
pid namespace is not it, IMO. Resource control (and other kinds of
task grouping behaviour) shouldn't require virtualization.

>
> a. Paul Menage's patches:
>
>         (tsk->containers->container[cpu_ctlr.subsys_id] - X)->cpu_limit

Additionally, if we allow mature container subsystems to have an id
declared in a global enum, then we can make the cpu_ctlr.subsys_id
into a constant.

>
> b. rcfs
>         tsk->nsproxy->ctlr_data[cpu_ctlr.subsys_id]->cpu_limit

So what's the '-X' that you're referring to

> 3. How are cpusets related to vserver/containers?
>
>         Should it be possible to, lets say, create exclusive cpusets and
>         attach containers to different cpusets?

Sounds reasonable.

>
> 6. As tasks move around namespaces/resource-classes, their
>    tsk->nsproxy/containers object will change. Do we simple create
>    a new nsproxy/containers object or optimize storage by searching
>    for one which matches the task's new requirements?

I think the latter.

>
>                 - If we don't support hierarchy in res controllers today
>                   but were to add that support later, then
>                   user-interface shouldn't change. That's why
>                   designining -atleast- the user interface to support
>                   hierarchy may make sense

Right - having support for a hierarchy in the API doesn't mean that
individual controllers have to support being in a hierarchy.

Paul