---------- Forwarded message ----------
Date: Fri, 18 Sep 1998 20:22:40 -0400
From: Guy T Almes <almes@internet2.edu>
To: Neal Cardwell <cardwell@cs.washington.edu>
Cc: Guy T Almes <almes@internet2.edu>,
Matt Zekauskas <zekauskas@advanced.org>
Subject: Re: Modeling of impact of loss on network performance
Neal,
I was very interested in the research you sketch below.
You might be interested in looking at some of the longitudinal
data we are developing for one-way delay and packet loss across
the university networks.
A summary can be view at:
http://ippm-db.advanced.org/plots
Understanding the relationship between end-to-end performance
and loss/delay would be very important to us, and the measurements
we're doing might be useful to you.
-- Guy
At 01:20 PM 8/25/98 -0700, Neal Cardwell wrote:
>Hi,
>
>There are several papers on modeling TCP performance in the presence of
>losses:
>
><LI> <B>[V97]</B> Curtis Villamizar,
><a href="ftp://engr.ans.net/pub/slides/nanog/feb-1997/tcp-congested.ps">
>TCP Response Under Loss Conditions</a>, slides from Feb 1997 NANOG</LI>
>
><LI>
><B>[LM97]</B>T.V.Lakshman and U.Madhow, <A
>HREF="http://tesla.csl.uiuc.edu/~madhow/publications/ton97.ps">The
>Performance of TCP/IP for Networks with High Bandwidth-Delay Products and
>Random Loss</A>, IEEE/ACM Trans. on Networking, June 1997</LI>
>
><LI>
><B>[MSMO97]</B> M. Mathis, J. Semke, J. Mahdavi, T. Ott, <A
>HREF="http://www.psc.edu/networking/papers/model_ccr97.ps">The
>Macroscopic Behavior of the TCP Congestion Avoidance Algorithm</A>, In
>Computer Communication Review, July 1997.</LI>
>
><LI> <B>[A98]</B> Anurag Kumar, <a href=""> Comparative Performance
>Analysis of Versions of TCP in a Local Network with a Lossy Link</a>,
>IEEE/ACM Transactions on Networking, Vol. 6, No. 4, pp. 485-498, August
>1998</LI>
>
><LI> <B>[PFTK98] </B>Jitendra Padhye, Victor Firoiu, Don Towsley, and Jim
>Kurose, <A HREF="http://www.acm.org/sigcomm/sigcomm98/tp/paper25.ps">
>Modeling TCP Throughput: A Simple Model and its Empirical Validation</A>,
>SIGCOMM '98, September, 1998.</LI>
>
>These are all looking at bulk transfers. A couple of us here at UW are
>looking at analyzing the performance of *short* TCP flows along paths
with
>random congestion-induced losses. We should have something by
mid-October.
>
>neal
>
>On Tue, 25 Aug 1998, Cindy Bickerstaff wrote:
>
>> Does any one know of people or groups of people (reference URLs?) who
are
>> doing modeling of expected performance in the presence of packet loss?
(e.g.
>> At thus and such loss level, connection setup time should be increased
by
>> blah.) How about accurately measuring loss these days when we know the
>> policies of many router owners include reduced priority for ICMP
packets? I
>> seem to remember the existance of people working on models but can't
find
>> any references pointing to equations or discussing the topic in depth.
>>
>>
>>
>>
>>
>> Cindy Bickerstaff
>> Internet Connectivity Engineering
>> Intel Corporation
>> cindy.bickerstaff@intel.com
>> (408)765-9857
>>
>>
>