<html>
BBN has recently completed a NASA funded research task on Explicit
Transport Error Notification (ETEN). <br>
The report and ns implementation are available at:<br>
<a href="http://wizards.ir.bbn.com/projects/pace/eten/" eudora="autourl">http://wizards.ir.bbn.com/projects/pace/eten/</a><br>
<br>
NASA funded this work because NASA would like to use COTS protocols such
as TCP. As many of NASA's links are error-prone, it would be
advantageous if TCP could determine the difference between a packet loss
due to errors and a packet loss due to congestion. This report is a
first pass at determining what the possible improvements may be if TCP
reacted to errors differently than to congestion. In addition,
issues related to practical deployment are discussed.<br>
<br>
In this study two types of ETEN mechanisms are proposed: (i)
per-packet<i> </i>mechanisms that notify endpoints<br>
of each detected corruption; and (ii) cumulative mechanisms that notify
endpoints of aggregate corruption<br>
statistics. The proposed mechanisms have been implemented in the
ns-2 simulator. The simulation<br>
results are presented on performance gains achievable for TCP Reno and
TCP SACK, using ETEN mechanisms over<br>
a wide range of bit error rates and traffic conditions. In addition, TCP
Reno and TCP SACK enhanced with<br>
ETEN mechanisms are compared against TCP Westwood, which uses a bandwidth
estimation strategy in place of the traditional<br>
AIMD congestion avoidance algorithm. Two issues related to the practical
deployment<br>
of ETEN mechanisms are discussed: corruption detection mechanisms (and
their co-operation with ETEN-based recovery<br>
in the transport layer) and security aspects.<br>
<br>
This work represents a very good start to investigations into ETEN.
I would appreciate any feedback related to the work. In particular,
thoughts on whether or not the potential performance improvements would
justify deployment. If deployment is justified, which areas of
research may be best to address? <br>
<br>
Thanks for your help.<br>
<br>
<br>
Will Ivancic<br>
<br>
<br>
<div>=======================================</div>
<div>Will Ivancic</div>
<div>NASA Glenn Research Center</div>
<div>21000 Brookpark Road MS 54-5</div>
<div>Cleveland, Ohio 44135</div>
<div>Phone +1 (216)433-3494</div>
<div>Fax +1 (216) 433-8705</div>
<div>Yahoo Instant Messenger ID: ivancic</div>
<div><a href="http://roland.grc.nasa.gov/~ivancic" EUDORA=AUTOURL>http://roland.grc.nasa.gov/~ivancic</a></div>
<br>
</html>