http://www.6assist.net/ - call for test

Max Tulyev maxtul at netassist.ua
Fri May 10 17:06:36 CEST 2013


Proto-41 or 6to4 is a point-to-point tunnels in any case. So if you want
to communicate directly to the significant part of the world this way,
you need to set up and maintain hundreds of tunnels, as well as hundreds
of BGP sessions. Also you have to go through a long and hard
administrative work to negotiate all those tunnels.

In our case, you need to set up only one tunnel and only one BGP session
to the ROUTE SERVER, which provide you all the routing information about
all the peers and prefixes.

On 10.05.13 17:11, Jeroen Massar wrote:
> On 2013-05-10 16:07 , Max Tulyev wrote:
>> Easy, just set up the BGP session using 6to4 address space ;)
>>
>> Still there are some major problems:
>> 1. Unlike IXP-like infrastructure, it is difficult to set up and
>> maintain a lot of BGP sessions ("each to each mesh").
> 
> Like an IXP infrastructure you would still have to configure these
> peers, and thus you will just have to automate it, which is something
> you want to do anyway if you want to scale in any matter or form...
> 
>> 2. The 6to4 infrastructure is unstable. The reachability of 6to4 address
>> is a far away from 100%.
>> 3. Your connectivity quality highly depends on 3rd party gate servers,
>> often has non-optimal paths from/to it, it can be overloaded, has a
>> packet loss and so on.
> 
> As your BGP peering will be between two 6to4 participants the packets
> will be sent directly between the 6to4 nodes as they know how to reach
> other.
> 
> The fun with using 6to4 addresses of course is that you for almost sure
> get stuck to a 1280 MTU which is very non-optimal.
> 
> 
> Better to then just use proto-41 tunnels and put the tunnel set up in
> the automated setup mentioned above.
> 
> Greets,
>  Jeroen
> 
> 




More information about the ipv6-ops mailing list