<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="Generator" CONTENT="MS Exchange Server version 6.5.7226.0">
<TITLE>RE: [e2e] performance of BIC-TCP, High-Speed-TCP, H-TCP etc</TITLE>
</HEAD>
<BODY>
<!-- Converted from text/plain format -->
<BR>
<P><FONT SIZE=2>Paul Francis wrote on Fri 2006-09-22 21:09<BR>
<BR>
> Any reason XCP not included in this?<BR>
<BR>
because XCP has morphed so many times no-one knows what it is anymore.<BR>
<BR>
XCP is no longer a clearly-defined algorithm -- it's now a brand name.<BR>
<BR>
L.<BR>
<BR>
<<A HREF="http://www.ee.surrey.ac.uk/Personal/L.Wood/">http://www.ee.surrey.ac.uk/Personal/L.Wood/</A>><L.Wood@surrey.ac.uk><BR>
<BR>
-----Original Message-----<BR>
From: end2end-interest-bounces@postel.org<BR>
[<A HREF="mailto:end2end-interest-bounces@postel.org">mailto:end2end-interest-bounces@postel.org</A>] On Behalf Of Douglas Leith<BR>
Sent: Friday, September 22, 2006 10:22 AM<BR>
To: end2end-interest@postel.org<BR>
Subject: [e2e] performance of BIC-TCP, High-Speed-TCP, H-TCP etc<BR>
<BR>
For those interested in TCP for high-speed environments, and perhaps also<BR>
people interested in TCP evaluation generally, I'd like to point you towards<BR>
the results of a detailed experimental study which are now available at:<BR>
<BR>
<A HREF="http://www.hamilton.ie/net/eval/ToNfinal.pdf">http://www.hamilton.ie/net/eval/ToNfinal.pdf</A><BR>
<BR>
This study consistently compares Scalable-TCP, HS-TCP, BIC-TCP, FAST-TCP and<BR>
H-TCP performance under a wide range of conditions including with mixes of<BR>
long and short-lived flows. This study has now been subject to peer review<BR>
(to hopefully give it some legitimacy) and is due to appear in the<BR>
Transactions on Networking.<BR>
<BR>
The conclusions (see summary below) seem especially topical as BIC-TCP is<BR>
currently widely deployed as the default algorithm in Linux.<BR>
<BR>
Comments appreciated. Our measurements are publicly available - on the web<BR>
or drop me a line if you'd like a copy.<BR>
<BR>
Summary:<BR>
In this paper we present experimental results evaluating the performance of<BR>
the Scalable-TCP, HS-TCP, BIC-TCP, FAST-TCP and H-TCP proposals in a series<BR>
of benchmark tests.<BR>
<BR>
We find that many recent proposals perform surprisingly poorly in even the<BR>
most simple test, namely achieving fairness between two competing flows in a<BR>
dumbbell topology with the same round-trip times and shared bottleneck link.<BR>
Specifically, both Scalable-TCP and FAST TCP exhibit very substantial<BR>
unfairness in this test.<BR>
<BR>
We also find that Scalable-TCP, HS-TCP and BIC-TCP induce significantly<BR>
greater RTT unfairness between competing flows with different round-trip<BR>
times. The unfairness can be an order of magnitude greater than that with<BR>
standard TCP and is such that flows with longer round-trip times can be<BR>
completely starved of bandwidth.<BR>
<BR>
While the TCP proposals studied are all successful at improving the link<BR>
utilisation in a relatively static environment with long-lived flows, in our<BR>
tests many of the proposals exhibit poor responsiveness to changing network<BR>
conditions. We observe that Scalable-TCP, HS-TCP and BIC-TCP can all suffer<BR>
from extremely slow (>100s) convergence times following the startup of a new<BR>
flow. We also observe that while FAST-TCP flows typically converge quickly<BR>
initially, flows may later diverge again to create significant and sustained<BR>
unfairness.<BR>
<BR>
--Doug<BR>
<BR>
Hamilton Institute<BR>
www.hamilton.ie<BR>
<BR>
<BR>
<BR>
<BR>
</FONT>
</P>
</BODY>
</HTML>