[e2e] TCP goodput as a function of connection count

[Lachlan Andrew]

Greetings Tim,

On 4 April 2010 04:45, Tim Dorcey <tim at ivisit.com> wrote:
> I have seen arguments in favor of the increased parallelism that distributed
> transports like bittorrent offer.  But, what does parallelism buy in a
> time-sharing environment?  I would expect nothing is gained by 10
> simultaneous file transfers each using 10% of 10 different paths, versus
> each using 100% of its own path.

If all 10 user shared exactly the same 10 paths, then your intuition
holds.  The theoretical benefit to multipath TCP is that each path is
typically shared with   different   numbers of users.  Each user can
explore multiple paths to find the ones which are shared with fewer
users, and send more data on those less-busy paths.  Thus, it is a
form of distributed traffic engineering.  Frank Kelly, Thomas Voice
and Peter Key have all done theoretical work in that area.

The practical benefits of multipath TCP are entirely different, like
robustness to disconnection.

On a quick scan, the "TCP connection game" paper seems to deal with
opening multiple connections over a single path.  The only benefits
that generally gives are (a) working around Reno's inability to fill
the path, and (b) providing a mechanism for unequal sharing of
capacity.  The (social) benefit from bittorrent is the multiple paths,
not just multiple connections.

Cheers,
Lachlan

-- 
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