[c-nsp] problem with a NPE-G2

[Frank Kempermann]

Hello to the list,

 

I currently have a 7204 NPE-400 with a GE-E controller running amongst
others. There are configured three full BGP feeds, as well as OSPF. I'm
using the latest 12.0 Service Provider IOS . This is running quite well,
however the machine is reaching its limits noticably now. A small DoS-attack
could breakdown the router.

 

My hardware dealer has given me a NPE-G2 for testing. According to Cisco it
should have twice the performance of an G1.

 

During the first tests with the same configuration as used with the NPE-400
I encountered the following problems which I wasn't able to solve by myself
so far.

 

A Ping to one of the IP-addresses configured on the Cisco GigE Interface
shows high latency every few pings; even higher as on the NPE-400...

 

rescue-ix:~# ping 193.xxxxxxxxx

PING 193.xxxxxxxxxxxxx (193.xxxxxxxxxxx) 56(84) bytes of data.

64 bytes from 193.xxxxxxxxxxx: icmp_seq=1 ttl=255 time=0.282 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=2 ttl=255 time=0.294 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=3 ttl=255 time=0.556 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=4 ttl=255 time=442 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=5 ttl=255 time=0.333 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=6 ttl=255 time=0.481 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=7 ttl=255 time=0.271 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=8 ttl=255 time=0.901 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=9 ttl=255 time=295 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=10 ttl=255 time=0.596 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=11 ttl=255 time=0.499 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=12 ttl=255 time=0.305 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=13 ttl=255 time=0.258 ms

64 bytes from 193.xxxxxxxxxxx: icmp_seq=14 ttl=255 time=149 ms

 

A "show process cpu" reads

16     4731200     21796     217067 16.47% 17.71% 17.65%   0 EnvMon

 

This is what seems to be eating up the CPU time... There's only a service
provider image available for the machine at this time. Because of that I'm
not able to test another IOS.

 

The utilization of the CPU is not high (20-30%), but why are the latencies
this bad when sending pings to the router?

 

I have completely disabled BGP, have completely disabled SNMP, the values
were unchanged.

 

When one is connected to the machine by telnet remotely, it seems there are
light "lags" from time to time. However, there are no peaks when looking at
"show process cpu history".

 

The routing through the router is 100% OK, bandwidth up to 300 MBit/s is
tested and poses no problems to the CPU. The only unanswered question is
where the lags and the sometimes high latency when pinging the router's IP's
is coming from.

 

Does anyone know this problem and maybe has a solution?

 

 

Many thanks in advance!

 

 

Regards, 

Frank