[e2e] common congestion controller for TCP connections

[Michael Savoric]

Hari Balakrishnan wrote:

Are you familiar with:

1) An Integrated Congestion Management Architecture
for Internet Hosts Hari 
Balakrishnan, Hariharan Rahul, and Srinivasan Seshan,
Proc. ACM SIGCOMM, 
Cambridge, MA, September 1999.  See 
http://nms.lcs.mit.edu/projects/cm/

2) RFC 3124

3) TCP Behavior of a Busy Internet Server: Analysis
and Improvements, Hari 
Balakrishnan, Venkata Padmanabhan, Srini Seshan, Mark
Stemm and Randy H. Katz, 
Proc. IEEE Infocom, San Francisco, CA, USA, March
1998.  (This seems to be a 
superset of your paper.)

> 
> Another reason I asked is that your paper does not seem to cite a lot of the
> related work nor address the points of difference clearly.  Doing so would be
> very useful.
> 
> Thanks!
> Hari

Thanks for your comments. Here are some remarks about related
approaches:

In
http://www-tkn.ee.tu-berlin.de/publications/papers/ccc_tr.pdf
I have described four related network information reuse or
common congestion control approaches: TCP Control Block
Interdependence (TCBI), Ensemble TCP (E-TCP), Shared Passive Network
Performance Discovery (SPAND), and the Congestion Manager (CM).

The main difference between EFCM and E-TCP is, that E-TCP wants
to control all TCP connections of an ensemble as aggressive to
the network as a single (!) TCP connection. EFCM controls an
ensemble of n TCP connections as aggressive to the network than
n standard TCP connections. In addition, EFCM does not need an
ensemble scheduler for sending TCP segments. And in contrast to E-TCP,
the EFCM controller jointly controls the timeout timer (SRTT, RTTVAR)
of TCP connections in an ensemble.
One advantage of E-TCP compared to EFCM is, that E-TCP reuses
network information between a new TCP connection and a recently
closed TCP connection (temporal TCBI). This feature will be implemented
in future versions of the EFCM controller.

The main difference between EFCM and CM is, that the EFCM controller
is transparent for all applications running on an end system.
Since I am not familiar with the details of the CM algorithms (are
there any documentations besides the source code available?) I cannot
compare EFCM and CM in this aspect.

Best regards,
Michael Savoric
   
======================================================
E-Mail:		savoric at ee.tu-berlin.de

Phone:		(+49 30) 314-23840

Fax:		(+49 30) 314-23818

Postal address:	Technical University Berlin
		Telecommunication Networks Group (TKN)
		Einsteinufer 25, 10587 Berlin
======================================================