Slow first traceroute time

[simon.wilson]

Hi,

I have 2 stacks of 3750s, which route between each other. If I tracert from a host on one site, i get this:

1 93 ms 193 ms 97 ms 10.1.8.1

2 <1 ms <1 ms <1 ms 172.16.1.1

3 <1 ms <1 ms <1 ms 10.1.0.35

But from the other site, this:

1 <1 ms <1 ms <1 ms 10.1.2.240

2 91 ms 95 ms 95 ms 172.16.1.2

3 <1 ms <1 ms <1 ms 10.1.8.52

Is the slow response from 172.16.1.2 (AKA 10.1.8.1) a problem?

Both switch stacks are broadly similar in config, although the slower switch is on 12.2(20).SE3, and the quicker one is on 12.2(25)SEB1

Thanks in advance.

[lgijssel]

What type of link is there between the two switches?

Looking at the delay-times makes me assume this is not direct ehernet.

Regards,

Leo

[simon.wilson]

This is 1Gb telco provided service across about 5Km.

Thanks

Simon

[lgijssel]

Looking at the delays it looks as if there is a problem with the link. You should get 1ms over this hop. Possibilities:

-Link problems (look for interface errors)

-Alternative routing. (you are using a different route)

Does this GB-link have the 172.16 adresses?

Regards,

Leo

[mheusinger · Accepted solution]

Hello,

creating ICMP packets is done in the switch CPU. With respect to other processes running this is a low priority task. So a better measure for your stack performance would be to use traffic through the switches and not to the switches.

The slow ICMP response does not mean that the switch throughput is slow or something is wrong. Those two types of traffic are handled quite different by the switches, so concluding from one type (f.e. ICMP from the switch) to the other type of traffic is not quite feasible.

In fact your traceroute shows exactly that: your third hop answers within 1 ms, so the packet had to be forwarded through the "slow" switch in less than this time.

In case your applications are doing well, I would not be worried.

Hope this helps! Please rate all posts.

Regards, Martin