|
On 2003-03-17T12:54:23,
Joseph Mack <mack.joseph@xxxxxxx> said:
> > I could write my own via perl to check things like load averages,
> no-one's done it (yet), presumably on the assumption that all machines
> are indentical and on the average will get identical loads.
I've actually done something like it in the past; I don't have the code any
longer, but it was a very simple "Poll the nodes every x seconds and recompute
weights based on load avg."; not exactly rocket science.
However, do keep in mind that loadavg is a poor indication of real resource
utilization; very much so, actually, but it might be enough. You should keep
in mind that a loadavg needs to be at least normalized via the number of CPUs
though ;-)
You may also find http://www.redfishsoftware.com.au/projects/feedbackd/ to be
of value to the task here.
Sincerely,
Lars Marowsky-Brée <lmb@xxxxxxx>
--
Principal Squirrel
SuSE Labs - Research & Development, SuSE Linux AG
"If anything can go wrong, it will." "Chance favors the prepared (mind)."
-- Capt. Edward A. Murphy -- Louis Pasteur
|
