[lwip-users] Snapshot of lwip based stack

[Kieran Mansley]

Hello All,

For the past year or so I've been building a stack for a high speed low
latency cluster network.  This used lwIP as a starting point, but has
steadily diverged and as such I'm unlikely to ever merge the changes I've
made back into the main tree.  I therefore thought I should at least make
some attempt to make the modifications I have made available in case
others are interested in them.

You can find a tar of my current tree at
http://www-lce.eng.cam.ac.uk/~kjm25/clan_lwip.tar.gz

I will do my best to keep this updated as I reach sensible release points.

The changes have been mainly aimed at getting high performance out of the
stack.  I'm not quite there yet, but can currently get hundreds of Mbps
using a benchmark called ttcp.  This is included in the above archive.

Please note, this comes with a number of caveats.
 - It is based on lwIP 0.5.2 (I think), although I have backported a few
of the more critical bug fixes that have been added to lwIP since.

 - The original lwIP code is copyright Swedish Institute of Computer
Science, and the author is Adam Dunkels.  All modifications have been made
by me, Kieran Mansley and are copyright Kieran Mansley, AT&T Research Ltd
and University of Cambridge.  Please observe this copyright and the
licence in each source file.

 - This code has been heavily tailored for the CLAN network.  This is a
high speed cluster network which exists only in my lab.  There has been no
attempt to make it portable or of use on any other hardware, and the
necessary libraries that are required to provide the network support are
not included.  As a result it will not compile.

 - This is just a snapshot of a work heavily in progress.  As it stands
most things that I use work enough for me to not get frustrated, but there
are a lot of part implemented and not implemented features.

 - There is little documentation, and the source probably only makes sense
to me.  It's best looked at after reading the paper I have written
describing my work, to be published at NICELI 2003 (part of ACM SIGCOMM).
I am currently reworking the paper to address the conference committee's
comments, but email me if you want a copy when that's done.  I am also
happy to answer intelligent questions if you're interested in getting a
more complete understanding.

Thanks!

Kieran