I have a problem with the distribution of bandwidth in HFSC queues.
We are using an OpenBSD 4.2 amd64 with two sk nics as a firewall/QoS
server. We use a lot o f HFSC queues with up to 3 levels of nesting.
OUR GOALS
We want to be SURE that all VoIP traffic is ALWAYS and IMMEDIATELY
forwarded. Then we want to distribute all the other available bandwidth
to the other queues based on different "weights" (and some other
restrictions).
OUR CURRENT IMPLEMENTATION
We use the "realtime" option only for a single queue where we put all
the VoIP traffic. All other queues are defined with different
"linkshare" options (the "bandwidth" is set to the same vale).
THE PROBLEM
From time to time, in saturation conditions, two sibling queues with
the SAME "linkshare" values receive completely different bandwidth: one
have ALL packets dropped, while the other have a lot a packets passed.
Why this happens?
Is this expected, acceptable?
Googling around I have found an email by "jared r r spiegel" which says:
in other words, a realtime declaraion specifies the bandwidth
that will be available if A) the queue wants it and possibly
B) saturation exists
if a queue does not have a realtime declaration, there exists
a possibilty that under saturation conditions, that queue
might not get much bandwith at all, possibly none.
It seems to imply that the situation described above is expected, even
if I cannot understand why!
Can anybody confirm this, and maybe explain why?
Thanks.
--
___________________________________________________
__
|- [EMAIL PROTECTED]
|ederico Giannici http://www.neomedia.it
___________________________________________________
