Cisco C1111-4P

[fpaulnsteiner]

We also use the router for bandwidth measurement before we hand over the service to our customer. We do this with

udp-small-services to echo the udp measurement traffic.
This works with this Routertype up to 35Mb/s,  higher bandwidths cause a 100% cpu load.

Alternatively, a Cisco 892 Router will provide 100 Mb/s in such a scenario.

Does anyone have any experience or an idea why a state of the art Router is moderately overwhelmed in cpu by this scenario?

Thanks

[balaji.bandi]

we like to know is your configuration, what causing the CPU high load?

these model sure need to give you more performance, but it was not, so we need to look in what causing this issue.

post below output.

show version

show  process cpu sorted

show run

[fpaulnsteiner]

Hello BB!

Cisco-1111#sh proc cpu sorted | e 0.00%
CPU utilization for five seconds: 79%/46%; one minute: 73%; five minutes: 39%
PID Runtime(ms) Invoked uSecs 5Sec 1Min 5Min TTY Process
508 67684 518849 130 32.87% 30.02% 15.80% 0 UDP Echo
121 765632 5935994 128 14.00% 14.36% 7.78% 0 IOSXE-RP Punt Se
9 2942648 650357 4524 0.47% 0.10% 0.06% 0 Check heaps
Cisco-1111#

Cisco-1111# sh interface GigabitEthernet0/0/0
...omitted
30 second input rate 39657000 bits/sec, 3552 packets/sec
30 second output rate 32315000 bits/sec, 2888 packets/sec
...omitted

Cisco-1111#sh version
...omitted
System image file is "bootflash:c1100-universalk9_ias.16.06.04.SPA.bin"
...omitted

Cisco-1111#sh run
Building configuration...

version 16.6
no service pad
service tcp-keepalives-in
service tcp-keepalives-out
service timestamps debug datetime msec localtime show-timezone
service timestamps log datetime msec localtime show-timezone
service password-encryption
service udp-small-servers
service sequence-numbers
platform qfp utilization monitor load 80
no platform punt-keepalive disable-kernel-core
!
hostname Cisco-1111#
!
boot-start-marker
boot-end-marker
!
!
no aaa new-model
clock timezone MEZ 1 0
clock summer-time MESZ recurring last Sun Mar 2:00 last Sun Oct 3:00
no ip source-route
!
!
!
!
!
!
!
no ip bootp server
no ip domain lookup
ip domain name a1.net
!
!
!
login on-success log
!
!
subscriber templating
!
!
!
!
vtp mode transparent
!
!
!
multilink bundle-name authenticated
!
!
!
vlan internal allocation policy ascending
no cdp run
!
!
!
!
interface GigabitEthernet0/0/0
description *** Provider WAN Interface ***
bandwidth 40000
ip address x.x.x.x 255.255.255.252
no ip redirects
no ip proxy-arp
load-interval 30
media-type rj45
negotiation auto
no cdp enable
!
interface GigabitEthernet0/0/1
description *** Customer LAN Interface ***
ip address x.x.x.x 255.255.255.240
no ip redirects
no ip proxy-arp
negotiation auto
no cdp enable
!
interface GigabitEthernet0/1/0
shutdown
!
interface GigabitEthernet0/1/1
shutdown
!
interface GigabitEthernet0/1/2
shutdown
!
interface GigabitEthernet0/1/3
shutdown
!
interface GigabitEthernet0/1/4
shutdown
!
interface GigabitEthernet0/1/5
shutdown
!
interface GigabitEthernet0/1/6
shutdown
!
interface GigabitEthernet0/1/7
shutdown
!
interface Vlan1
no ip address
!
...omitted

Cisco-1111#

interesting logg during test:
000683: May 24 08:00:14.124 MESZ: %IOSXE-5-PLATFORM: R0/0: cpp_cp: QFP:0.0 Thread:000 TS:00004561030145814691 %PUNT_INJECT-5-DROP_PUNT_CAUSE: punt cause policer drop packet casue 11

I have tested with command configured:

platform punt-keepalive disable-kernel-core

its the same cause.

It looks like there is a build in cpp!

best regards

[Giuseppe Larosa]

Hello fpaulnsteiner,

if there is an auto CoPP applied on the device this can also influence the results of your performance test.

UDP traffic may be rate limited at some point.

Even if I agree that the UDP echo process looks like to use a lot of CPU

Cisco-1111#sh proc cpu sorted | e 0.00%
CPU utilization for five seconds: 79%/46%; one minute: 73%; five minutes: 39%
PID Runtime(ms) Invoked uSecs 5Sec 1Min 5Min TTY Process
>> 508 67684 518849 130 32.87% 30.02% 15.80% 0 UDP Echo

see CoPP best practices

https://www.cisco.com/c/en/us/about/security-center/copp-best-practices.html

Hope to help

Giuseppe

[balaji.bandi]

adding to other comment, worth looking this bug also :

https://bst.cloudapps.cisco.com/bugsearch/bug/CSCvj62552/?rfs=iqvred

[fpaulnsteiner]

Hello BB, sorry for the delay!

The root cause for this problem seems to be the punt-policer, ( COPP is not configured) but when i disable it, there is no difference in the bandwith measurement.
Cisco-1111(config)# platform punt-policer 11 ?             
  <10-64600>  max punt rate for cause For-us data
  alerting    Enable punt-policer's notification

Cisco-1111(config)# platform punt-policer 11 64600 ?     
  <1-226100>  max burst size for cause For-us data
  high        High priority traffic only
  <cr>
Cisco-1111(config)#no platform punt-policer 11 64600 226100

"%PUNT_INJECT-5-DROP_PUNT_CAUSE: punt cause policer drop packet casue 11"
Does anyone have an idea how to really switch off the punt policer?

Best regards