[c-nsp] MLPPP

[Rodney Dunn]

On Thu, Mar 26, 2009 at 10:30:08AM -0400, Jason Berenson wrote:
> Rodney,
> 
> It's running: 12.4(18a).  I had to downgrade from the latest about 6 
> months ago because of a bug where 'show policy' would show no output 
> even if QoS was working properly.
> 
> router#show int mul2 stat
> Multilink2
>          Switching path    Pkts In   Chars In   Pkts Out  Chars Out
>               Processor          0          0          0          0
>             Route cache      18049    3982931      25553   14234069
>                   Total      18049    3982931      25553   14234069
> router#show int mul2 stat
> Multilink2
>          Switching path    Pkts In   Chars In   Pkts Out  Chars Out
>               Processor          0          0          0          0
>             Route cache      18601    4110973      26553   14682852
>                   Total      18601    4110973      26553   14682852
> 
> fonseca#show cef in mul 2
> Multilink2 is up (if_number 132)
>  Corresponding hwidb fast_if_number 132
>  Corresponding hwidb firstsw->if_number 132
>  Internet address is 10.3.4.229/30
>  ICMP redirects are always sent
>  Per packet load-sharing is disabled
>  IP unicast RPF check is disabled
>  Inbound access list is not set
>  Outbound access list is not set
>  Interface is marked as point to point interface
>  Hardware idb is Multilink2
>  Fast switching type 7, interface type 105
>  IP CEF switching enabled
>  IP CEF VPN Feature Fast switching turbo vector
>  IP Null turbo vector
>  VPN Forwarding table "nypirg"
>  Input fast flags 0x1000, Input fast flags2 0x0, Output fast flags 
> 0x4000, Output fast flags2 0x0
>  ifindex 127(127)
>  Slot -1 Slot unit 2 Unit 2 VC -1
>  Transmit limit accumulator 0x0 (0x0)
>  IP MTU 1500
> 
> Does that mean that there's no processor switching going on there? 

Yep. It's all being interrupt switched so you should be fine.

> Why 
> would a VRF make any difference to the MLPPP?

forwarding vectors are different. But in this code we have the hooks
to do MPLSoMLPPP if that's what you were doing..which you are not.
The vrf interface on a bundle isn't what we call MPLSoMLPPP...that's when
you enable MPLS on the bundle.

  I see the same outputs 
> for a non VRF'd MLPPP.

It's working as it should.

With the new PA the overall CPU would be less b/c the mlppp work is offloaded
to an asic on the PA.

> 
> -Jason
> 
> Rodney Dunn wrote:
> >You have it in a VRF which really shouldn't cause an issue as it's
> >tag2ip and ip2tag.
> >
> >What code is it?
> >
> >Make sure it's the latest 12.4 mainline as we did some work in 12.4 to make
> >this work.
> >
> >Can you get a 'sh int mul 2 stat' after a clear counters...get it a few
> >times and send it?
> >
> >Also, what are the other interface configs feeding this bundle?
> >It could be features on them causing the punts.
> >
> >What does 'sh cef int' say?
> >
> >
> >Rodney
> >
> >On Wed, Mar 25, 2009 at 05:03:03PM -0400, Jason Berenson wrote:
> >  
> >>Here's a sample:
> >>
> >>interface Multilink2
> >>ip vrf forwarding VPN1
> >>ip address x.x.x.x 255.255.255.252
> >>no cdp enable
> >>ppp multilink
> >>ppp multilink group 2
> >>service-policy output voice
> >>!
> >>interface Serial6/0/25:0
> >>no ip address
> >>encapsulation ppp
> >>down-when-looped
> >>no cdp enable
> >>ppp multilink
> >>ppp multilink group 2
> >>!
> >>interface Serial6/0/26:0
> >>no ip address
> >>encapsulation ppp
> >>down-when-looped
> >>no cdp enable
> >>ppp multilink
> >>ppp multilink group 2
> >>!
> >>
> >>-Jason
> >>
> >>
> >>Rodney Dunn wrote:
> >>    
> >>>The G1's with MLPPP should not be process switching the traffic.
> >>>
> >>>What is the config?
> >>>
> >>>The EC cards just offload the MLPPP to the new asic on the PA.
> >>>
> >>>Rodney
> >>>
> >>>On Wed, Mar 25, 2009 at 04:35:50PM -0400, Jason Berenson wrote:
> >>> 
> >>>      
> >>>>Greetings,
> >>>>
> >>>>I've got a 7206VXR NPE-G1 with a bunch of DS3 cards in it (PA-MC-T3).  
> >>>>There's about 25 multilinks with an average of 2 T1s per bundle.  I see 
> >>>>a lot of process switching on the router and I have a feeling it's 
> >>>>because we don't have the PA-MC-T3-EC card so the processor has to step 
> >>>>in for the MLPPP. 
> >>>>
> >>>>Is this the case?  If I get some PA-MC-T3-EC cards to swap in, will 
> >>>>that take a lot of load off the NPE-G1?  Any output needed, please let 
> >>>>me know.
> >>>>
> >>>>Thanks,
> >>>>Jason
> >>>>_______________________________________________
> >>>>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/
> >>>>   
> >>>>