[altq 867] Measuring packet delays

[Nicolas Christin <nicolas@cs.virginia.edu>]

	Hi all,

	I'm pretty new to ALTQ, which, by the way, is an outstanding
package. I was wondering if somebody had already had a wonderful idea for
measuring packet delays (please don't tell me I've got to hack into the
IP header, it's not an option for me...).

	The only thing I found in the archive was [altq 336] and it was
not that helpful, unfortunately. What I'm doing at this point is to
enqueue a timestamp in a list each time a packet comes in, and to dequeue
the timestamp when a packet leaves. I have two questions:

	1) can I be guaranteed that there will not be any synchronization
problems between packet arrivals/departures and timestamp
enqueueing/dequeueing?  I.e., if mydiscipline_enqueue() contains a
"timestamp_enqueue()" call, and if mydiscipline_dequeue() contains a
"timestamp_dequeue()" call, is it clear that the timestamps
enqueued/dequeued will exactly match the packets enqueued/dequeued? I
believe FreeBSD does not support kernel-level threads (is that correct?),
which alleviates a possible cause for such synchronization problems, but
what about Open/NetBSD? Are there other possible causes for
synchronization problems I need to worry about when using such a timestamp
mechanism?

	2) does anybody know of any external package that does just that
(measuring individual packet delays) so that I can compare the results
with what my method gives? The results I'm getting from my external
timestamp list seem to be right, but I'd like to know if there's a way
to double-check.

	Any help is appreciated. Thanks in advance!

	Best regards,
	--Nick

------------------------------------------------------------------------
Nicolas Christin                        nicolas@cs.virginia.edu
Multimedia Networks Group               http://www.cs.virginia.edu/~nc2y
------------------------------------------------------------------------