[j-nsp] IPv6 best practice

Truman Boyes truman at suspicious.org
Wed May 27 03:40:07 EDT 2009


When you say parallel IPv6 network do you mean on the same  
infrastructure links that you use for IPv4? You can certainly do this.

If you want tunnel IPv6 over IPv4 with GRE you can do this with JUNOS:


This may be useful if you want to have some customers or peers to  
connect with but your upstream is not routing IPv6 at this time.

What JUNOS does not do at this time is 6to4 or NAT-PT. At this time,  
dual stack is probably the way to go for (in your example FTTH)  
subscribers. The IETF and SP community is still trying to sort out  
what to do when we truly run out of IPv4 addresses and we *can't* do  
dual stack because we only have IPv6. Carrier Grade NAT (CGN)  
solutions have been proposed, and they may get us out of trouble for  
sometime, however NAT'ing millions of sessions is going to be very  
intensive and will take time to perfect.


On 26/05/2009, at 4:35 PM, Matthias Gelbhardt wrote:

> Hi!
> Perhaps the great advantage we have, is that our network is rather  
> small at the moment. So this might be the perfect time to start the  
> use of IPv6 apart from starting to using it right from the beginning.
> In the meantime I have run a few tests, and it seems to be very easy  
> to set up a parallel IPv6 network.
> Anyway, we are just starting to understand the concepts, so  
> apologize for any "dumb questions" I have. But here is another one:  
> I read about, that several tunnel techniques do not work in juniper.  
> Is it today possible, to give a customer only IPv6 access (for  
> example a FTTH link), to let him access the whole internet? Are  
> there any conversions available on Juniper, that can do that?
> Maybe I first should get our prefix announced, before getting to the  
> complex questions ;)
> Regards,
> Matthias
> Am 26.05.2009 um 07:17 schrieb Truman Boyes:
>> Hi,
>> Congrats! If you have MPLS in your backbone, you can continue to  
>> use IPv4 as the transport for your MPLS signaling. With this  
>> approach you can run 6VPE and build a VPN for your inet6 traffic.  
>> This is a common approach for "getting things going". All v6 stuff  
>> just rides across MPLS and your core doesn't know the difference.
>> If you plan to carry inet6 routes in a non-ipvpn topology, you can  
>> get your core "dual stack"; if you run OSPFv2 today you will need  
>> to consider OSPFv3 or ISIS to carry IPv6 in your IGP.
>> You can have all your ipv4 configuration coexist with ipv6.
>> You can also use logical routers / virtual routers / etc, if you  
>> want separation at that level.
>> Kind regards,
>> Truman
>> On 25/05/2009, at 10:32 PM, Matthias Gelbhardt wrote:
>>> Hi!
>>> Today we received our IPv6 prefix from RIPE. There are many  
>>> projects in the pipeline, so we would like to enable our network  
>>> with IPv6 and would like to start new projects with default  
>>> deployment in IPv6.
>>> What would be the best practice to bring our backbone to IPv6?  
>>> Would one use a virtual router for this? Or is it doable to  
>>> configure IPv6 in parallel to my other configuration? Do I have  
>>> special memory requirements for a virtual router, like a VM on  
>>> VMware?
>>> Regards,
>>> Matthias
>>> _______________________________________________
>>> juniper-nsp mailing list juniper-nsp at puck.nether.net
>>> https://puck.nether.net/mailman/listinfo/juniper-nsp

More information about the juniper-nsp mailing list