[e2e] Inferring loss at Client-side on TCP

David G. Andersen dga at lcs.mit.edu
Mon Mar 3 21:04:57 PST 2003


On Tue, Mar 04, 2003 at 02:56:32AM -0000, Kalyan quacked:
> Hello,
> 
> I am a grad student working on developing a tool which should 
> capture the loss based statistics at the client-side. The possible 
> inputs to my tool are tcpdump traces.
> 
> Can anyone suggest some heuristics of identifying loss based on 
> client-side tcpdump traces.

Depends on the fidelity with which you want to identify them.  You
could always look for triple dup acks and assume whatever losses
you haven't seen.  Assuming you want something more accurate,
Crovella and Barford's Critical Path Discovery techniques would
seem a reasonable starting point.  They present an off-line algorithm
for matching TCP packets with the ACK that released them, which has
obvious application to identifying loss.  However, if your long-term
goal is to do this detection online, it may not be applicable.

  -Dave

-- 
work: dga at lcs.mit.edu                          me:  dga at pobox.com
      MIT Laboratory for Computer Science           http://www.angio.net/
      I do not accept unsolicited commercial email.  Do not spam me.




More information about the end2end-interest mailing list