Re: [RRG] Are we solving the wrong problem?

["Mark Handley" <M.Handley@cs.ucl.ac.uk>]

On Feb 19, 2008 4:28 PM, David Williamson <dlw+rrg@tellme.com> wrote:
> A simple approach might be to try to stripe UDP traffic across all
> available paths.  I hope it is obvious why this is a bad idea with
> real-time traffic.

For UDP, clearly a lot depends on the application.  In general, L4
multihoming/mobility should work well for UDP too, so long as the
flows are congestion controlled, and so long as the delay of the
slowest path is acceptable.

For non-congestion controlled traffic, there's no control loop to
react to congestion and cause the apps to move the data to the other
paths.  You can probably get crude balancing, but there's nothing to
reduce the overload on the congested paths.  Likely any crude
congestion control would be sufficient to get benefits though if the
flows are long-lived.

For real-time traffic, the real question is one of delay, not
reordering.  Such apps need a playout buffer anyway to remove jitter,
and this should remedy reordering too (if it doesn't already do this,
various wireless links will badly affect your apps).  For streaming
media, there's no real problem, but for interactive real-time apps,
the effects of the jitter buffer are effectively to increase the delay
to that of the slowest path.  The likely solution would be to have a
delay threshold, and only use paths that are less than that threshold.
 However, there are a lot of unanswered questions about the stability
of such solutions.  Only research will tell what the best solutions
here are.

Cheers,
Mark

--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg