[c-nsp] DTS traffic shaping & voice queuing

Rodney Dunn rodunn at cisco.com
Tue Dec 21 13:31:35 EST 2004


Very good example.  It's kinda hard to turn down the
clock rate on a FE connection that someone wants to
provide subrate speeds on though.

If the rate at which you can clock
out bits in the first Tc with shaping exceeds the policing configuration
on the downstream next hop (with or without burst configured)
then that does create a problem where the downstream could
drop.  The ability for the device to send at exactly the
same rate constantly is a function of the granularity
at which the tokens are added. 

I personally have never worked on a setup where this
was the issue but in theory I can see it happening if
the policing granularity is more granular than the shaping
granularity and/or the burst sizes don't match up.

This may not apply to the deployment scenario that was
being asked about though.

But good point.

Rodney



On Tue, Dec 21, 2004 at 11:17:31AM -0700, Clinton Work wrote:
> 
> I have found clocking down a circuit provides much better shaping than
> using a MQC traffic shaper. The problem with the software based shaper 
> is that it doesn't smooth out bursts. Tokens (bytes) are put into
> the bucket at 10ms intervals, but the entire bucket can be used up in 
> the first 1ms  by a large burst. IP packets after the token is exhausted 
> will be queued until token bucket is refreshed (in 9ms). If a downstream
> device is policing traffic to the shaped rate with a small burst size
> then the bursts will be dropped.
> 
> The software driver for a interface will smooth out bursts as packets 
> are taken out of the queue and transmitted at a regular interval.
> 
> Token refresh rate (10 - 125ms).
> http://www.cisco.com/en/US/tech/tk543/tk545/technologies_tech_note09186a00800a3a25.shtml#tokenrefreshrate
> 
> Rodney Dunn wrote:
> > That's not really true.  I think of it this way.  If you have a T1
> > you can only clock bits out on the wire at a T1 rate.  So the shaping
> > is /*if you will*/ built in to the circuit.  Therefore to do QOS you
> > just apply the queueing policy on the interface.


More information about the cisco-nsp mailing list