Hello,
this is my first post here and it is about the behavior of a HFSC class
with a two piece-wise linear service curve.
My goal is to separate audio streams from bulk traffic. To keep it simple
a use two classes one high priority realtime and the other for the rest.
Example:
tc qdisc add dev ppp0 root handle 1: hfsc default 11
tc class add dev ppp0 parent 1: classid 1:1 hfsc rt m1 400kbit d 30ms m2
100kbit ls rate 100kbit ul rate 500kbit
tc class add dev ppp0 parent 1: classid 1:11 hfsc ls rate 400kbit ul rate
500kbit
[filter...]
As I understand it the first piece (high slope) of the service curve in
class 1:1 effects the first deadline of packets in the class. It shortens
the time before packets leave the class for the first time. Later on the
timeline the 2nd piece of the service curve has a constant lower slope
(100kbit). and defines the deadlines'' intervals (when packets leave the
class). This intervals are constant and higher than the first
"manipulated" deadline. As a result of the shortened first deadline
all
later deadline time points are nearer the arriving time of its packets
which gives as a lower delay. (typically they also arrive in canstant
intervals considering audio or video streams)
Now coming to my question I am wondering if this "high slope effect"
of
the first service curve piece only takes effect for the first time the
class is used (first 10ms). If this is the case only the first time a
audio stream flow through class 1:1 it''s delay will be lowered. Later
starting flows will not profit of the high m1 slope. To be honest I think
it is just a bit of a wrong understanding by me here, isn''t it?
Regards,
Uwe
--
Using Opera''s revolutionary e-mail client: http://www.opera.com/mail/
