[rbak-nsp] SE800 QoS profiles

[Sridhar]

Hi Jim,
When you mention that traffic to the user goes over an L2TP tunnel, I
assume you are referring to the SE800 functioning as an LNS.

It is possible to configure a QoS queuing policy on the subscriber
profile either through Radius or local configuration on the SE800,
that prioritizes downstream traffic to the user, while simultaneously
metering it as well. This configuration does not require that the
SE800 know the actual DSL sync speed, but assumes that it will be at
least whatever the user is paying for  (8Mb or 2Mb in your example).

Here's what the config would look like with a 2Mb limit. Using a
marking policy at the ingress port where the traffic comes in from the
core, VoIP traffic goes into queue 1 below, and non-VoIP into queue 2

qos policy l2tp-silver pwfq
 rate maximum 2000
 num-queues 4
 queue 0 priority 0
 queue 1 priority 1
 queue 2 priority 2 weight 100
 queue 3 priority 2 weight 50
 queue priority-group 0 rate 128
 queue priority-group 1 rate 512
 queue priority-group 2 rate 1500 exceed

You can see that:
1. the policy caps the aggregate traffic at 2000Kbps
2. VoIP traffic in queue 1 has strict higher priority than non-VoIP
traffic in queue 2
3. Priority-group 1 (queue 1) traffic is capped at 512 Kbps
4. Priority-group 2 (queue 2) traffic is capped at 1500 Kbps, but
allowed to exceed this cap if there is no higher priority traffic.

Hope this helps.

thanks
sridhar


On Fri, Sep 4, 2009 at 7:30 AM, Jim Tyrrell<jim at scusting.com> wrote:
> Hi,
>
> I want to create and apply some QoS profiles so that protocols such as VOIP
> get priority on certain users connections.  ie If the user has an 8Mb
> connection then VOIP should get priority over all their other traffic up to
> say 6Mb.  But if the user had a 2Mb line we may want VOIP to get priority up
> to 1.5Mb against the users other traffic.
>
> Our users have DSL connections and sync anywhere between 512k and 24Mb but
> we dont necessarily  know what all the users sync speeds are.  So my
> question is, do we need to know what the end users actual line speed is in
> order to apply QoS and prioritise their various protocols?  I  was told by a
> Redback engineer that you dont need to know the end user line speed as you
> can just use rate percentage in the qos profile?
> Is this right as I dont see how the SE800 could prioritise traffic without
> knowing what the upper limit is?  The SE800 is sending the traffic to the
> user in an L2TP tunnel though a GigE card so surely it would need to know
> the user has a line rate of 8Mb so that when overall traffic for the
> specific user starts hitting that limit it needs to start dropping/queue the
> lower priority packets?  Otherwise as far as the Redback is concerned
> wouldnt the upper limit be the GigE port?
>
> Thanks.
>
> Jim.
> _______________________________________________
> redback-nsp mailing list
> redback-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/redback-nsp
>