[e2e] traffic engineering considered harmful

[Jon Crowcroft]

In message <Pine.GSO.4.32.0106131232370.14210-100000 at caruaru>, "DJamel H. Sadok
" typed:

 >>I think the idea is good, but there are problems (nothing new!)

the idea was partly prompted by an e-mail from Jonathan Smith at
upenn...so i can't claim its mine!
 
 >>What sort of contracts are we going to have if clients have
 >>control over networks that today is the business of ISPs? who is to
 >>blame when things go wrong? who pays for bad engineering?
 
well, in the M3I project, we assume that we translate congestion on a
route into a price people pay to the owners of the resources o nthat
path  - now this DOES assume a large scale deployment of loits of
brokers (we don't pay routers, we pay brokers who pay ISPs - the
brokers also accumulate statistics so that people can make informed
choices)

 >>One basic type of control we would like to see is a way for a custumer to
 >>measure network performance without having to deploy tons of application
 >>and/or network level processes at varios points in the network. Most of
 >>the time you only have control/access over your access point and don't
 >>even know what the egress node of your traffic is!

a fully distributed (only end system ) model is feasible but hard to
get right...

 >>I had  a look at the RON approach and it is based on some
 >>kind of overlay network  to get performance/routing information. Is there
 >>a simpler way of doing  these things? I need to make sure that my ISP is
 >>obeying our contract and giving me the real picture and that I am paying
 >>for real traffic!

i think sall these overlay trhings start out as pure end system, then
to get scaling they put in some sort of (self organisign) hierarchy of
"distinguished" servers that can give peopel more stable views of
things...
 >>
 >>On Wed, 13 Jun 2001, David P. Reed wrote:
 >>
 >>> At 09:42 PM 6/12/01 +0000, Bob Braden wrote:
 >>>
 >>> >   *>
 >>> >   *>    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
 >>> >   *>
 >>> >
 >>> >Why isn't this the Tragedy of the Commons waiting to happen?
 >>>
 >>>
 >>>
 >>> Bob - to me, this RON idea is economic arbitrage.  The more likely "tragedy
 >>> of the commons" arises because of the selfish behavior of the AS's in
 >>> seeking to create bottlenecks to extract unjustified rents, and the
 >>> arbitrage reduces its likelihood.
 >>>
 >>>
 >>> - David
 >>> --------------------------------------------
 >>> WWW Page: http://www.reed.com/dpr.html
 >>>
 >>>
 >>>
 >>

 cheers

   jon