[c-nsp] "Packets dropped to the next slow path"

[Rodney Dunn]

On Tue, Nov 22, 2005 at 12:19:18PM +0100, Vincent De Keyzer wrote:
> Right Rodney, that's exactly what we are doing (and BTW, how to configure
> this was explained to me two years ago on this excellent list).
> 
> So basically, my problem is that CEF does not support PPP half-bridging, so
> if I want to offer VoIP to this customer, I have two options:
> 1. move him to a router where CPU usage is not subject to peaks (which is
> what I did temporarily)

I wouldn't want to be relying on process switching to service delay/jitter
sensitive traffic. Process scheduling will become an issue to service the
packets.
 
s> 2. install an E1 router (instead of a bridge) at customer premises, and do
> plain PPP

Yeah. That's the best option is to route out to the customer side.
> 
> Please confirm my understanding...
> 
> Vincent
> 
> PS: "sh int stat" indeed shows that traffic is switched by the processor
> mostly (why not completely? Seems like 0.2% of the packets do find their way
> via CEF, weird!)

I've never messed with it. I might try to set it up in the lab for another
issue I was looking at with it but I'm really backed up right now so it
may be a bit.

> 
> 
> > I'd never seen that either until last week when someone complained
> > that their counters were not working for MQC policing.
> > 
> > ppp bridge ip is like a BVI under the covers to bridge
> > ethernet frames out over the PPP link from what I see.
> > 
> > 
> > 
> > se2/0
> > ip add 1.1.1.1  --- pppp link ---- remote router --- ethernet
> > 
> > and the remote router is like a transparent bridge between
> > the ip of 1.1.1.1 to the remote 1.1.1.0/24 subnet.
> > 
> > I have no idea if we support that in anything other than process
> > level code. I've never tested it.
> > 
> > Rodney
> > 
> > >
> > > I've never seen anyone assign a MAC address to a serial interface.
> > What's
> > > the purpose of that, the secondary addresses, and "ppp bridge ip"?
> > >
> > > I kind of wonder if the fact that you seem to be enabling bridging on
> > the
> > > interface is why the traffic isn't cef switched...though have you
> > verified
> > > it's really not by doing a show int stats?  That'll probably show
> > most/all
> > > the traffic is process switched.
> > >
> > >
> > > ----------------------------------------------------------------------
> > >   Jon Lewis                   |  I route
> > >   Senior Network Engineer     |  therefore you are
> > >   Atlantic Net                |
> > > _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
> > > _______________________________________________
> > > cisco-nsp mailing list  cisco-nsp at puck.nether.net
> > > https://puck.nether.net/mailman/listinfo/cisco-nsp
> > > archive at http://puck.nether.net/pipermail/cisco-nsp/
> > _______________________________________________
> > cisco-nsp mailing list  cisco-nsp at puck.nether.net
> > https://puck.nether.net/mailman/listinfo/cisco-nsp
> > archive at http://puck.nether.net/pipermail/cisco-nsp/