[Tmrg] Total number of TCP modifications (variants)?

lachlan.andrew at gmail.com (Lachlan Andrew) Wed, 17 February 2010 21:04 UTC

From: lachlan.andrew at gmail.com (Lachlan Andrew)
Date: Thu, 18 Feb 2010 08:04:13 +1100
Subject: [Tmrg] Total number of TCP modifications (variants)?
In-Reply-To: <201002171914.30671.v13@v13.gr>
References: <201002171914.30671.v13@v13.gr>
Message-ID: <aa7d2c6d1002171304q4c6f4a86r50e5a44cc475df45@mail.gmail.com>

Greetings Alexsander,

As Stefanos pointed out, there are very many congestion control
algorithms for TCP.

Most of those listed on Wikipedia don't actually change the protocol
("bits on the wire"), just the underlying algorithm for calculating
the congestion window.  If you include ones which change the protocol,
there are hundreds.  Most are in the academic literature, rather than
the IETF/IRTF world.

The TCPM working group keep making modifications (like appropriate
byte counting) without calling them "new versions".  Many of these
modifications can be used in conjunction with one another (giving an
exponential number of variants), but some of them conflict.

Michael Welzl and Wes Eddy have assembled a list of RFCs relating to
congestion control.
<http://tools.ietf.org/rfcdiff?url2=http://tools.ietf.org/id/draft-irtf-iccrg-cc-rfcs-07.txt>
 (It is in the process of becoming an RFC itself.)

Why do you want to know the number?  If you want to learn the
literature generally, the best option is to choose a particular
weakness of Reno, and look at proposals to address that particular
weakness.  Examples are
 - poor performance at high bandwidth-delay products
 - poor performance in the presence of packet loss not due to congestion
 - lack of priority (there is current interest in low-priority variants of TCP)
 - unfairness to flows with different round trip times
 - filling (possibly large) buffers at bottleneck links, causing excessive delay

I hope that helps,
Lachlan

On 18 February 2010 04:14, Stefanos Harhalakis <v13 at v13.gr> wrote:
> Hello,
>
> On Wednesday 17 of February 2010, Aleksandar Milenkoski wrote:
>> First of all, thank you for the prompt answers for my previous question.
>> What bothers me now is, does anyone has an idea which is the total (or
>> ?approx. total) number of so-far published TCP modifications (like Reno,
>> ?Vegas..)? I did not manage to find any official TCP modification list.
>> If anyone has some info about this problem, please inform me.
>
> Linux kernel (as of 2.6.32) includes 13 congestion control algorithms.
> Wikipedia article [1] lists 22 including Vista's Compound-TCP. I'd guess that
> it's in the range of 22-28, but this isn't a much-informed guess.
>
> [1] http://en.wikipedia.org/wiki/TCP_congestion_avoidance_algorithm
> __________________________________llack of ack of _____________
> Tmrg-interest mailing list
> Tmrg-interest at ICSI.Berkeley.EDU
> http://mailman.ICSI.Berkeley.EDU/mailman/listinfo/tmrg-interest
>



-- 
Lachlan Andrew  Centre for Advanced Internet Architectures (CAIA)
Swinburne University of Technology, Melbourne, Australia
<http://caia.swin.edu.au/cv/landrew> <http://netlab.caltech.edu/lachlan>
Ph +61 3 9214 4837