Joseph W. Doherty Wed, 10/15/2008 - 04:18

I haven't seen that well documented. Likely reason, most modern switches seem to have unloaded latencies under 1 ms, and actual latencies most dependent on internal queuing based on traffic load and traffic distribution. (Probably similar reason why you no longer see "fragment free" or "cut thru" switching to try to reduce switch latency.) You mind find this documented for especially high performance switches designed for supporting something like grid computing.


This Discussion