From: Sam Vilain (sam_at_vilain.net)
Date: Thu 15 Aug 2002 - 18:56:59 BST
Jacques Gelinas <jack_at_solucorp.qc.ca> wrote:
> > Is there any work going on regarding limit the memory and cpu of each
> > vserver?
> Yes and no. The various limit are the next thing on the list, but
> work has not started (not that I know).
I'm working on CPU scaling with 2.4.19-acX - implemented as a `CPU
token bucket' system, where each s_context get N cpu tokens every X
jiffies, and your s_context's process priority is reduced or increased
if you have few or many CPU tokens in your bucket.
So it won't be CPU limiting or minimum cycles guaranteeing, but you
will be able to say that on average over busy periods (where each
vserver is contributing at least 1 to the load) vserver X gets 50% of
the CPU when it's busy, vserver Y gets 25% and vserver Q gets 25%,
etc.
This will apply to the O(1) scheduler, too 8-). I have implemented it
but am having problems with kernel memory allocation proving too hard
for me right now. RSN :-)
--
Sam Vilain, sam_at_vilain.net WWW: http://sam.vilain.net/
7D74 2A09 B2D3 C30F F78E GPG: http://sam.vilain.net/sam.asc
278A A425 30A9 05B5 2F13
Real software engineers like C's structured constructs, but they are
suspicious of it because they have heard that it lets you get "close
to the machine."