Re: small stuff

[Olivier Bonaventure <Bonaventure@info.ucl.ac.be>]

Iljitsch,

> Then there are some other issues that I don't want to spend too much text on:

> 
> RTTs
> 
> 
> We had a discussion on whether measuring round trip times and basing decisions on the measurements is useful. I think it is. 

I also believe that delay is an important criteria in the selection of
the IPv6 addresses to be used for a communication.

> Delay is easy to measure, and all things being equal, a path with a long delay is less desireable than a path with a long delay. As I wrote last week, the queuing delay part in end-to-end delay is getting less and less important as bandwidth goes up.
> 
> And one of the main goals of being smart about reachability detection is avoid the really bad paths. As such continuing the reachability detection a bit longer if the delay looks worse than the previously looking path is an easy way to do this in many cases.
> 
> The argument that we shouldn't duplicate transport work doesn't convince me as we're lower in the stack and as such we get to see things (locator info) that is invisible to TCP and the like.

One issue with selecting the best local and distant addresses is that it
could required a large number of rtt measurements if done naively. This
is especially important when sites have more than two providers. In
today's Internet, multihomed stub ASes can have between 2 and 25 providers.

However, selecting the best pair of IPv6 addresses does not necessarily
require a large number of rtt measurements. In the peer-to-peer
literature, several papers have proposed coordinate systems. Basically,
the idea is that each host performs rtt measurements with a small number
of distant hosts (say a dozen). Based on those rtt measurements, each
host computes its coordinates (say (x,y)). The nice property of those
coordinates is that if you attach a coordinate to each IPv6 address,
then the distance between two coordinates is a good approximation of the
rtt between the two hosts. This means that by performing a few dozens of
rtt measurements, a host can predict the rtt towards *any* distant host.

Cédric de Launois developped such a coordinate system for his PhD
thesis. In his thesis, he shows that the coordinate system is stable and
the evaluation performed based on measurements obtained from the RIPE
Test Traffic project show that the rtt estimation is accurate. By using
a coordinate system, it is possible to avoid really bad paths and in
most cases, the path selected is the best one or very close to the best one.

You can find Cédric's thesis (Unleashing Traffic Engineering for IPv6
Multihomed Sites) on the web :
http://www2.info.ucl.ac.be/people/delaunoi/publications.html

Cédric also wrote an internet draft describing how coordinates could be
embedded in the DNS :
http://www2.info.ucl.ac.be/people/delaunoi/ietf/draft-de-launois-dnsext-sloc-rr-00.txt

I think that using a coordinate system to aid in the selection of
addresses, even as an option, would be a very insteresting addition to
the SHIM6 solution that is currently being developped.


Best regards,


Olivier Bonaventure


-- 
CSE Dept. UCL, Belgium - http://www.info.ucl.ac.be/people/OBO/