There are two other considerations: 1. ICMP packets might follow a different path than the application in the presence of ECMP 2. The ICMP responder might rate limit and drop if it's a router regardless of the drop characteristics of the path -- RFC 6192.
Thanks, Thumb typed by Carlos Pignataro. Excuze typofraphicak errows On Aug 6, 2013, at 4:33 PM, "Anoop Ghanwani" <an...@alumni.duke.edu<mailto:an...@alumni.duke.edu>> wrote: On Mon, Aug 5, 2013 at 8:54 PM, Fan, Peng <fanp...@chinamobile.com<mailto:fanp...@chinamobile.com>> wrote: A basic requirement for using ICMP (and other active measurement approaches) is to let probing traffic and service traffic have the same drop probability, though it is usually difficult to guarantee this. It seems like this would be further complicated if something like ECN is in use. While ECT packets would be marked with CE and thus not contribute to the loss rate, ICMP packets would end up being discarded. Anoop _______________________________________________ OPSAWG mailing list OPSAWG@ietf.org<mailto:OPSAWG@ietf.org> https://www.ietf.org/mailman/listinfo/opsawg
_______________________________________________ OPSAWG mailing list OPSAWG@ietf.org https://www.ietf.org/mailman/listinfo/opsawg