Time required to shut down a single interface on Catalyst 2950

Hi all,

I have been doing some experiments trying to determine the time required to shut down a single FastEthernet interface on a Catalyst 2950.

In these experiments, I have been connecting to the switch from a Linux box serial port to the console port on the switch. I am using the standard cable that came with the switch, and am operating at 57600bps with hardware flow control enabled. (I know the switch technically supports

115200bps, but I experienced problems at this rate).

One simple approach was to measure the time difference between when the 'sh' command was given and when the prompt was returned. Using this approach over several thousand runs, the average was about ~23ms or so.

It has been argued that perhaps the interface is effectively shut down long before the prompt is returned. Some have even argued that the port is shut down within the first couple milliseconds, and the rest of the time is spent "cleaning up" after the action.

Unfortunately, measuring time accurately at such fine granularity is quite difficult. I've not yet been satisfied with any of my experiments to definitively prove (within ~1ms accuracy) how fast an interface can be effectively shut down.

I was hoping one of two things may result from this post:

1) Someone may already know the answer, or can argue for or against the results/methodologies mentioned above

2) Someone can think of a clever experiment to actually produce results within ~1ms accuracy

Thanks, Adam