Concurrent Multipath Transfer (was Re: [Tsvwg] sctp performance)

Hi Brian/all,

(I trimmed the Cc list. And I apologize for posting on several lists. But
this discussion started on all these lists. Maybe it could move to
one/more lists as per interest?)

The papers that are referred to in this thread are:
http://www.cis.udel.edu/~iyengar/publications/2004.spects.iyengar.pdf
http://www.cis.udel.edu/~iyengar/publications/2004.icon.iyengar.pdf

I add another paper here, which is currently under submission and is a
continuation of CMT work:
http://www.cis.udel.edu/~iyengar/publications/2005.icdcs.iyengar.pdf

> As I'm sure you know, the OpenSS7 SCTP implementation has been using a
> weighted version of the algorithms from the second paper since 2001.

[...]

> The implementation also uses a (patentable) FR algorithm that is far
> simpler than the split-FR presented in the first paper.

I certainly did not know that you already implemented these algorithms. I
don't think any of the co-authors on the papers knew either. Could you
post the algos, or send a reference that we could look up?

If your FR algo is far simpler, I would certainly like to look at it,
because I could not get the SFR-CACC algo to be simpler than it is. Well,
it is simpler than it _was_, but I do not know how to make it any simpler.
Also, the first paper presents two other algos - the CUC and DAC algos. At
least the CUC algorithm is, IMHO, necessary to see expected throughput, at
least if the paths in question have different end-to-end delays. Did  you
implement something like CUC too?

You say the algo is "patentable" - I do not understand what that means.
Did you mean to say the algo is "patentED"? So, can you post the algo?

Note: The ID (draft-iyengar-sctp-cacc-xx.txt) which was first published in
Feb 2002 has expired now, but I'm planning to resubmit it in a new form
and with significant additions/modifications. I wanted to spend more time
looking at changeover/CMT concerns with SCTP before putting something out
as an ID.

regards,
jana

---------------------------------------------------------------
Janardhan R. Iyengar           http://www.cis.udel.edu/~iyengar
Protocol Engineering Lab  --   CIS   --  University Of Delaware
---------------------------------------------------------------

_______________________________________________
tsvwg mailing list
tsvwg@ietf.org
https://www1.ietf.org/mailman/listinfo/tsvwg