[e2e] packet-pair probe implementation

[Constantine Dovrolis]

Atsuo,

there is a lot of recent research on bandwidth estimation, and on
the packet pair technique in particular. We recently wrote
a survey paper (currently under submission) on bandwidth
estimation that gives all (?) the recent references:

http://www.cc.gatech.edu/~dovrolis/Papers/bwest_survey.pdf

To answer your question about packet pairs "in the real world":

Even though the packet pair idea is very simple, it is
quite challenging to get an accurate capacity estimate
from it in practice. The reason is not only because
"the real world is filled with noise and dirt", but because
our original understanding and theories of how the packet
pair technique works were quite limited. In particular,
they were ignoring the major effect that cross traffic
has on packet pairs in FCFS queues. Please see the references
of the previous paper for details.

To make the long story short, sending just a few packet pairs and
averaging the resulting measurements will not work, especially in
paths that carry some non-negligible traffic, and even
more in high-bandwidth paths. Some of the recent works
in this area are based on elaborate statistical techniques,
and on filtering of the measurements based on the queueuing
effects that "shape" the packet pairs as they move through
the network. My friend Andre Broido from CAIDA describes
these techniques as "network spectroscopy", I think for a very good
reason.

For a measurement tool that uses packet pairs to estimate
the end-to-end capacity of a path, you can look at Pathrate
(available at www.pathrate.org).

Constantinos

--------------------------------------------------------------
Constantinos Dovrolis | 218 GCATT | 404-385-4205
Assistant Professor | Networking and Telecommunications Group
College of Computing | Georgia Institute of Technology
dovrolis at cc.gatech.edu
http://www.cc.gatech.edu/fac/Constantinos.Dovrolis/

On Thu, 1 May 2003, Atsuo J. wrote:

> Is the packet-pair probe (bandwidth estimation) technique (introduced by Keshav and revised by Paxson) really useful in the realistic network?Does anybody test it on the real testbed? Also, are there any real implementation/usage of this technique at all? Your suggestions will be very appreciated. Atsuo
>
> ---------------------------------
> Do you Yahoo!?
> The New Yahoo! Search - Faster. Easier. Bingo.