> > > There are few problems if we aggressively cleanup the lru lists.
> > >   - we may be throwing the pages out which may be pulled in
> > >       by the task in very near future.
> >
> > This is only the case if the processes within the same class
> > approach or exceed the memlimit.
>
> Wouldn't that be the most often case ?

In our system we have on the order of 200 classes.  Most of these classes
stay well within their memory limit.  There are a few that push the memory
limit, and it is those that I'd prefer to penalize.

>
> >
> > >   - if a class need lot more than the limit, we will be
> > >       swapping bigtime to the disk, which would slow down
> > >       the other processes in the system
> >
> > Why?  Because the class is consuming too much I/O bandwidth?
> Well, that's
>
> yes, that is why...
> > what a working I/O controller should solve, right?
>
> In the absence of a I/O controller, other classes would be
> affected for no fault of their own..

Agreed... which is why CKRM needs a working I/O controller (either Shailab's
or the folks working on Cello).

Marc



-------------------------------------------------------
This SF.Net email is sponsored by: New Crystal Reports XI.
Version 11 adds new functionality designed to reduce time involved in
creating, integrating, and deploying reporting solutions. Free runtime info,
new features, or free trial, at: http://www.businessobjects.com/devxi/728
_______________________________________________
ckrm-tech mailing list
https://lists.sourceforge.net/lists/listinfo/ckrm-tech

Reply via email to