From: Sam Vilain (sam_at_vilain.net)
Date: Wed 11 Jun 2003 - 17:21:24 BST
On Wed, 11 Jun 2003 14:45, Herbert Poetzl wrote:
> I do not know much about advanced scheduling (yet),
> but it sounds like a good approach to me, although
> I neither had time to test your patches, nor did
> read anything about results/improvements ...
> anyway, I recently coded some 'virtual' abstraction
> (see my posting on the vserver mailing list for
> further details), which as far as I can tell, would
> benefit from a scheduler which actually knows about
> the virtual contexts, and honors per virtual cpu-
> time limitations and priorities, etc, while acting
> like the conventional scheduler within one virtual.
> what do you think about this?
> any hints/ideas/suggestions?
about the best thing I could suggest would be to look at the original
patch I posted, as this is just the incremental changes to the vserver
kernel for the TBF filter. The TBF algorithm itself is quite simple.
It's at http://vilain.net/linux/patch-2.4.21-pre4-ac6+ctx16-o1tbf
(warning to `passers-by': don't actually use that patch for a running
system, use the ones in http://vilain.net/linux/ctx/ instead)
I would suggest that you read the patch, try to understand how it
works and try to implement it in your changed version. I am happy to
answer whatever questions you have about how it works, and I will
certainly make an effort to review any implementation that you knock
together at the source level.
If you aren't using a kernel based on the O(1) scheduler already, this
will be nigh on impossible to port plainly; or, rather, a different
strategy will need to be developed. What kernel tree are you basing
your efforts on?
-- Sam Vilain, sam_at_vilain.net
Software suppliers are trying to make their software packages more user-friendly [....] Their best approach, so far, has been to take all the old brochures, and stamp the words, 'user-friendly' on the cover. BILL GATES