[e2e] traffic engineering considered harmful

[David G. Andersen]

Jon Crowcroft just mooed:
> 
> Why I think traffic engineering is a Bad Thing
> 	by jon crowcroft
> 
> [...]
> 
> what would be better? what would be better would be to give the users
> a choice - if loose source routing (or strict, or scalelble
> multihoming, or IPv6 GSE) worked well, then users would choose the
> best path for their traffic. if a user who sees a poor packet loss
> (or ECN marking) rate on path a can try a different path, then a market 
> in alternate paths would develope - this would obviate the need for
> traffic engineering, and would make it an end2end selection (hence why
> i am posting this here)

   I'll steal this topic as a chance for some blatant self-promotion:

   Resilient Overlay Networks:    http://nms.lcs.mit.edu/ron/

   Take a small collection of hosts around the 'Net.  They
can see different paths in and out of various ASs.  Have them
measure the paths between each other, and if they can establish
a better route by sending their packets indirectly through another
member of the overlay, do so.

   It's a rough approximation of the ideal that you're alluding to
in your message, since it has the obvious downsides of needing to
go all the way to the edge and then back in, and it's limited in its
view of the available paths, but it's one way to start doing some
of the things you're looking at.  Works pretty well, too, especially
in the face of a few egregiously bad links.

   -Dave

--
work: dga at lcs.mit.edu                          me:  dga at pobox.com
      MIT Laboratory for Computer Science           http://www.angio.net/