[c-nsp] high cpu usage because of process switching on mlppp interface

Bruce Pinsky bep at whack.org
Wed Aug 29 13:08:24 EDT 2007


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Arda Balkanay wrote:
> Hi
> i have a problem about high cpu usage on a 7206 VXR NPE-G1 because of
> process switching on a multilink ppp interface.
> 
> I have two E1 interfaces between 7206 and 3845 (below MTP-C72-02 indicates
> 7206 and ADN-C38-01 indicates 3845).
> I've bundled these two E1 interfaces. And when the amount of traffic rised
> CPU usage also increased about 15 percent more.
> "show proc cpu" output points increase at the processes of ip input and tag
> input.
> MTP-C72-02#sh proc cpu sorted
> CPU utilization for five seconds: 78%/64%; one minute: 79%; five minutes:
> 81%
>  PID Runtime(ms)   Invoked      uSecs   5Sec   1Min   5Min TTY Process
>   68    39242000 258183793        151  8.92%  8.20%  8.40%   0 IP
> Input
>  109    25093708 119681214        209  6.88%  7.37%  7.79%   0 Tag Input
> 
> 
> When i looked at the switching status of the interface i saw that 7206 is
> using process switching while 3845 uses fast switchig (especially for mpls)
> MTP-C72-02#sh inter mul 1 switching
> Multilink1 Maltepe - Adana MLPPP
>           Throttle count          0
>                    Drops         RP     752707         SP          0
>              SPD Flushes       Fast          0        SSE          0
>              SPD Aggress       Fast          0
>             SPD Priority     Inputs        259      Drops          0
> 
>     Protocol  IP
>           Switching path    Pkts In   Chars In   Pkts Out  Chars Out
>                  Process   50148189  395174672     261949   26235771
>             Cache misses          0          -          -          -
>                     Fast       1325     105874      12194     724886
>                Auton/SSE          0          0          0          0
> 
>     Protocol  Tag
>           Switching path    Pkts In   Chars In   Pkts Out  Chars Out
>                  Process   79390997 3601460527       5116     538398
>             Cache misses          0          -          -          -
>                     Fast          0          0   87354606 4071850482
>                Auton/SSE          0          0          0          0
> 
> 
> ADN-C38-01#sh inter mu 1 switching
> Multilink1 Adana - Maltepe MLPPP
>           Throttle count          0
>                    Drops         RP          0         SP          0
>              SPD Flushes       Fast          0        SSE          0
>              SPD Aggress       Fast          0
>             SPD Priority     Inputs        242      Drops          0
> 
>     Protocol  IP
>           Switching path    Pkts In   Chars In   Pkts Out  Chars Out
>                  Process      32394    2110804      17595    1260300
>             Cache misses          0          -          -          -
>                     Fast     559376   60411948        154       8439
>                Auton/SSE          0          0          0          0
> 
>     Protocol  Tag
>           Switching path    Pkts In   Chars In   Pkts Out  Chars Out
>                  Process       7106     398899      10784     539203
>             Cache misses          0          -          -          -
>                     Fast   87130476 3140243335   80454345  488241675
>                Auton/SSE          0          0          0          0
> 
> 
> When i checked "show ip interface" output i can see that fast swithing is
> enabled but i didn't understand why 7206 uses process switching while 3845
> uses fast
> swithcing with the same multilink interface configuration. ip interface
> output is as follows:
> 
> MTP-C72-02#sh ip int mu 1
> Multilink1 is up, line protocol is up
> ...
>   IP fast switching is enabled
>   IP fast switching on the same interface is enabled
>   IP Flow switching is disabled
>   IP CEF switching is enabled
>   IP CEF Fast switching turbo vector
>   IP Null turbo vector
>   IP multicast fast switching is enabled
>   IP multicast distributed fast switching is disabled
>   IP route-cache flags are Fast, CEF
>   ...
> MTP-C72-02#
> 
> ADN-C38-01#sh ip int mu 1
> Multilink1 is up, line protocol is up
>   ...
>   IP fast switching is enabled
>   IP fast switching on the same interface is enabled
>   IP Flow switching is disabled
>   IP CEF switching is enabled
>   IP CEF Fast switching turbo vector
>   IP multicast fast switching is enabled
>   IP multicast distributed fast switching is disabled
>   IP route-cache flags are Fast, CEF
>   ...
> ADN-C38-01#
> 7206 uses ios c7200-jk9s-mz.123-14.T3.bin and 3845 uses
> c3845-spservicesk9-mz.124-8b.bin
> 
> 
> Do you have any idea about this situation ?
> 
> 


Take a look at the output of "show cef interface" which will tell you why
packets would be dropped into a slower switching path:

Dialer1 is up (if_number 6)
  Corresponding hwidb fast_if_number 6
  Corresponding hwidb firstsw->if_number 6
  Internet address is 172.25.1.1/24
  ICMP redirects are never 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 on input is disabled
  BGP based policy accounting on output is disabled
  Interface is marked as point to point interface
  IPv4 packets switched to this interface are dropped to the next slow
path: Dialer
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  Hardware idb is Dialer1
  Fast switching type 15, interface type 96
  IP CEF switching enabled
  IP CEF switching turbo vector
  IP prefix lookup IPv4 mtrie 8-8-8-8 optimized
  Input fast flags 0x0, Output fast flags 0x0
  ifindex 5(5)
  Slot -1 Slot unit -1 VC -1

- --
=========
bep

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFG1agIE1XcgMgrtyYRAi6QAJ0VZo1dbjmwdvk8lSFj7TqzzfjnOwCfYjNs
MXUi5c2Q+MAkn+vzJtWzcWg=
=sRhw
-----END PGP SIGNATURE-----


More information about the cisco-nsp mailing list