[Devel] Re: [ckrm-tech] [RFC][PATCH 2/7] UBC: core (structures, API)

[Srivatsa Vaddagiri]

On Tue, Aug 22, 2006 at 04:46:50PM +0400, Kirill Korotaev wrote:
> Srivatsa Vaddagiri wrote:
> > On Mon, Aug 21, 2006 at 03:02:17PM +0400, Kirill Korotaev wrote:
> > 
> >>>Except that you eventually have to lock ub0. Seems that the cache line
> >>>for that spinlock could bounce quite a bit in such a hot path.
> >>
> >>do you mean by ub0 host system ub which we call ub0
> >>or you mean a top ub?
> > 
> > 
> > If this were used for pure resource management purpose (w/o containers)
> > then the top ub would be ub0 right? "How bad would the contention on the
> > ub0->lock be then" is I guess Matt's question.
> Probably we still misunderstand here each other.
> top ub can be any UB. it's children do account resources
> to the whole chain of UBs to the top parent.
> 
> i.e. ub0 is not a tree root.

Hmm ..if I understand you correctly, there is no one single root of the
ubc tree? In other words, there can be several roots (each representing
a distinct group of processes)? CKRM has one single root afaik, under
which multiple resource/task groups are derived.

-- 
Regards,
vatsa