[ipv6-wg] Youtube over IPv6!

On 11 feb 2010, at 15:31, Peter van Eijk wrote:

> A couple of days ago Martin Milnert wrote
> 
> <qoute>
> The time when it began is clearly visible in our own IPv6 graphs,
> http://stats.csbnet.se/public/ipv6/csbnet-ipv6-traffictypes.html ,
> and you can see similar correlated data at for example
> http://www.de-cix.net/content/network.html .
> </quote>
> 
> What puzzles me is that we see none of this at the AMS-IX, have a look
> at http://www.ams-ix.net/sflow-stats/ipv6/
> 
> Does anybody have a clue as to why that is?
> 
> Peter van Eijk
> (dutch IPv6 taskforce)
> 


I can think of various reasons:

- scale

The CBS graph show an increase of a few megabit, that won't show up in graphs at gigabit level.

- It only works for selected parties

It's hard to tell which networks actually have their nameservers whitelisted by google, I know a few who do...I can probably also find a number of bigger ones in NL who don't.

- Routing preference

A lot of traffic is based around a few tunnel operators and another part is 6to4 anycast. It might just be that route selection happens to decide the path over DECIX is preferred to a path over Amsterdam. It might also be that Amsterdam traffic happens to be on a private interconnect instead of the shared medium.

- Local situation

We know France has a relatively high number of IPv6 users, I have seen sources which indicate the same goes for Russia. It might be that those parties involved are only situated in Frankfurt and not in Amsterdam.

And I'm not going to open the can of worms labeled 'sampling interval'.

So graphs and especially the public ones don't necessarely reflect reality, the only way to be 100% sure is to get to the graphs on the end points. We normally don't publish statistics and I won't do now, but I can tell you the impact of youtube got lost in the day to day lump of usenet trraffic we see passing on our interfaces.

MarcoH