Re: [ckrm-tech] [patch00/05]: Containers(V2)- Introduction [message #6747 is a reply to message #6652] |
Thu, 21 September 2006 00:34  |
Paul Menage
Messages: 642 Registered: September 2006
|
Senior Member |
|
|
On 9/20/06, Chandra Seetharaman <sekharan@us.ibm.com> wrote:
>
> What I am wondering is that whether the tight coupling of rg and cpuset
> (into a container data structure) is ok.
Can you suggest a realistic scenario in which it's not? Don't forget
that since the container abstraction is hierarchical, you don't have
to use both at the same level. So you could easily e.g. have a parent
container in which you bound to a set of memory/cpu nodes, but had no
rg limits, and several subcontainers where you configured nothing
special for cpuset parameters (so inherited the parent params) but
tweaked different rg parameters.
Paul
|
|
|