Mailing List Archive: CBWFQ via radius for l2tp?

CBWFQ via radius for l2tp?

May 25, 2005, 9:01 AM

Post #1 of 5 (3335 views)

I'm trying to figure out a way to configure QoS for output traffic to DSL
customers we see as L2TP PPP sessions from the ILEC. The setup uses
vpdn-groups to handle all customers for each group/area with
virtual-template1. I tried building a service-policy on the aggregation
router and applying it via an lcp:interface-config cisco-avpair, but then
the test user could not reconnect. With debugging, I saw:

May 25 11:43:03: Vi100 VTEMPLATE: Messages from (un)cloning ...
Class Based Weighted Fair Queueing will be applied only to the
Virtual-Access interfaces associated with an MLP bundle.

Is there any way to get CBWFQ without changing the virtual-template to
multilink (which I'm worried may break everything, since nobody is
actually doing multilink)?

The router is a 3640 running 12.2M code with a 4 port T1 IMA card.

----------------------------------------------------------------------
Jon Lewis | I route
Senior Network Engineer | therefore you are
Atlantic Net |
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________

Re: CBWFQ via radius for l2tp? [ In reply to ]

dpeng at cisco

May 25, 2005, 10:10 AM

Post #2 of 5 (3211 views)

Permalink

We currently do not support queueing on any type of VPDN vaccess
interface (though this will change in the somewhat near future). You
can do rate-limiting or policing.

Dennis

Jon Lewis [jlewis@lewis.org] wrote:
> I'm trying to figure out a way to configure QoS for output traffic to DSL
> customers we see as L2TP PPP sessions from the ILEC. The setup uses
> vpdn-groups to handle all customers for each group/area with
> virtual-template1. I tried building a service-policy on the aggregation
> router and applying it via an lcp:interface-config cisco-avpair, but then
> the test user could not reconnect. With debugging, I saw:
>
> May 25 11:43:03: Vi100 VTEMPLATE: Messages from (un)cloning ...
> Class Based Weighted Fair Queueing will be applied only to the
> Virtual-Access interfaces associated with an MLP bundle.
>
> Is there any way to get CBWFQ without changing the virtual-template to
> multilink (which I'm worried may break everything, since nobody is
> actually doing multilink)?
>
> The router is a 3640 running 12.2M code with a 4 port T1 IMA card.
>
> ----------------------------------------------------------------------
> Jon Lewis | I route
> Senior Network Engineer | therefore you are
> Atlantic Net |
> _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
> _______________________________________________
> cisco-nas mailing list
> cisco-nas@puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nas

Re: CBWFQ via radius for l2tp? [ In reply to ]

jlewis at lewis

May 25, 2005, 10:14 AM

Post #3 of 5 (3226 views)

Permalink

On Wed, 25 May 2005, Dennis Peng wrote:

> We currently do not support queueing on any type of VPDN vaccess
> interface (though this will change in the somewhat near future). You
> can do rate-limiting or policing.

That kind of sucks for now. So you're suggesting I can use CAR or
policing in a service policy to prevent !voip traffic from using more than
C-V bandwidth, where C is the circuit's capacity, and V is what we need
reserved for voip packets?

Any idea how far off (somewhere in 12.3, 12.4, 13.x) CBWFQ over VPDN is?

----------------------------------------------------------------------
Jon Lewis | I route
Senior Network Engineer | therefore you are
Atlantic Net |
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________

Re: CBWFQ via radius for l2tp? [ In reply to ]

dpeng at cisco

May 25, 2005, 10:39 AM

Post #4 of 5 (3206 views)

Permalink

Jon Lewis [jlewis@lewis.org] wrote:
> On Wed, 25 May 2005, Dennis Peng wrote:
>
> > We currently do not support queueing on any type of VPDN vaccess
> > interface (though this will change in the somewhat near future). You
> > can do rate-limiting or policing.
>
> That kind of sucks for now. So you're suggesting I can use CAR or
> policing in a service policy to prevent !voip traffic from using more than
> C-V bandwidth, where C is the circuit's capacity, and V is what we need
> reserved for voip packets?

Yeah, I think that is the only way. The other concern with doing QoS
on the LNS is that usually at least one segment of the path to the
client is oversubscribed (usually the uplink on the DSLAM), so you
can't really do end-to-end QoS.

> Any idea how far off (somewhere in 12.3, 12.4, 13.x) CBWFQ over VPDN is?

It'll come out in 12.2SB, the broadband release off of 12.2S and I
believe we will only support the 7x00, 1000x platforms in that
release. Scalability will also decrease due to the increased memory
requirements for doing shaping/queueing on those interfaces.

Dennis

> ----------------------------------------------------------------------
> Jon Lewis | I route
> Senior Network Engineer | therefore you are
> Atlantic Net |
> _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________

RE: CBWFQ via radius for l2tp? [ In reply to ]

mm at elabnet

May 31, 2005, 2:34 PM

Post #5 of 5 (3229 views)

Permalink

> Jon Lewis [jlewis@lewis.org] wrote:
> > On Wed, 25 May 2005, Dennis Peng wrote:
> >
> > > We currently do not support queueing on any type of VPDN vaccess
> > > interface (though this will change in the somewhat near
> future). You
> > > can do rate-limiting or policing.

This statement blows me quite away, that's a quite hard thing.. I never needed it (or was too lazy to find out in detail) up to now, but ever thought I could do CBWFQ if I need to. Now I know at least why it hasn't worked at all when I tried..

> > That kind of sucks for now.
yes, it really does ;)
I wonder what all the currently popping-up VoIP-DSL-pricedumpers here do.. I guess they simply don't do any QoS at all.

> Yeah, I think that is the only way. The other concern with doing QoS
> on the LNS is that usually at least one segment of the path to the
> client is oversubscribed (usually the uplink on the DSLAM), so you
> can't really do end-to-end QoS.
now I haven't seen any oversubscription so far here, I guess this is also provider/country dependent.
The PPPoE sessions I get over L2TP should have 1:1 bandwidth down to the subscriber..

> > Any idea how far off (somewhere in 12.3, 12.4, 13.x) CBWFQ
> over VPDN is?
>
> It'll come out in 12.2SB, the broadband release off of 12.2S and I
> believe we will only support the 7x00, 1000x platforms in that
> release. Scalability will also decrease due to the increased memory
> requirements for doing shaping/queueing on those interfaces.

As I never understood (and probably never will) what the various SXY releases are good for, will this also go into any "mainline" I can wait for like 12.4T ?

Michael