[c-nsp] MLPPP and ip load-sharing per-packet

Jon Lewis jlewis at lewis.org
Mon Oct 10 15:25:51 EDT 2005


On Mon, 10 Oct 2005, Rodney Dunn wrote:

> Check 'sh cef in' for the bundle interface and see if it tells you anything.
> You have to check the features on the downstream interfaces to see
> if you have a feature on that will force a punt in that code.

Nothing real interesting.  Downstream interfaces (I assume you mean the 
interfaces input traffic from the mu1 interface will be switched to) are 
DS3s doing nothing special (tag-switching...but we do that all over the 
network).

Multilink1 is up (if_number 77)
   Corresponding hwidb fast_if_number 77
   Corresponding hwidb firstsw->if_number 77
   Internet address is 209.208.2.5/30
   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
   IP policy routing is disabled
   BGP based policy accounting is disabled
   Interface is marked as point to point interface
   Hardware idb is Multilink1
   Fast switching type 7, interface type 95
   IP CEF switching enabled
   IP CEF Flow Fast switching turbo vector
   IP Null turbo vector
   Input fast flags 0x0, Output fast flags 0x4000
   ifindex 69(69)
   Slot -1 Slot unit 1 VC -1
   Transmit limit accumulator 0x0 (0x0)
   IP MTU 1500

#sh int mu1 stat
Multilink1
           Switching path    Pkts In   Chars In   Pkts Out  Chars Out
                Processor    1587703  346958159       1136     126270
              Route cache        515      50302    1852816  891831028
                    Total    1588218  347008461    1853952  891957298

I realized I'd left tag-switching enabled on the member T1s.  I removed it 
from them, so it's only on the mu1 interface.  Right after doing that, 
there was a burst of route-cache processed input packets...but within a 
few seconds, it was back to process switching all but the occasional 
packet.

interface Multilink1
  load-interval 30
  tag-switching mtu 1520
  tag-switching ip
  service-policy output mu1out
  no cdp enable
  ppp authorization noauth
  ppp multilink
  no ppp multilink fragmentation
  multilink-group 1
  hold-queue 200 in
  hold-queue 200 out

interface Serial5/0/1:0
  bandwidth 1536
  no ip address
  encapsulation ppp
  load-interval 30
  no fair-queue
  ppp authorization noauth
  ppp multilink
  multilink-group 1

6 more Serial's configured identically.

I found that without the manually bumped up hold-queue's I had around 10% 
packet loss.  I suspect they won't be necessary once the traffic is being 
properly cef switched.

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


More information about the cisco-nsp mailing list