[afnog] The path to eliminating IPv4 in the backbone

Andrew Alston Andrew.Alston at liquidtelecom.com
Mon Oct 2 09:24:24 UTC 2017


Hi All,

I figured I’d drop a note that may be of interest to some of you.

Some background – Large networks run MPLS – that – is a fact of life in the modern world – love it or hate it.

Now, thanks to what I personally believe amounts to dysfunction on behalf of both the IETF and the Vendors – this has always been a problem – because so much of what we do with MPLS is still so tied to IPv4.  The good news though – over the last few days I’ve been doing some substantial testing on a variety of technologies – and progress is being made.

So here is what we still cannot do:

You cannot bind a martini cross connect to v6 addresses on either a cisco or a juniper – despite ldpv6 finally being available
You cannot do labelled unicast on a v6 neighbour on a Cisco – you are limited to v6 labelled unicast over v4 neighbours (used for 6pe stuff)
You cannot do l2vpn address family on v6 neighbours yet

However – here is what you *can* do:

Create a traffic engineering tunnel between two v4 end-points with all the intermediate routers doing V6 – this is done using SRTE over either node-sid’s or adj-sid’s.  It works because traffic is entirely address family agnostic once it leaves the initial router – and is purely switched on the labels.  So for example if I create a path that looks like this:

explicit-path name te-sr255
index 1 next-label 800101
index 2 next-label 806208
index 5 next-label 806311
index 6 next-label 809635
index 7 next-label 800435
!

Where Index 1 is a v4 Node-SID, index 2 5 and 6 are all V6 Node-SID’s and index 7 is a V4 SID – and I then bind a pw-class to that – I can force a pseudowire to transverse over v6 only sections of the network.

There are still some limitations to this – because if you have a need to do circuit by circuit TE – the functionality is not there yet in Cisco and while there is code out there to do it in Juniper it’s not yet available in general releases (60 days odd?)  I’ve also sent a query to Cisco to find out when exactly they will be supporting this.  Juniper also introduced the ability to dynamically create v4ov6 tunnels in 17.3 – meaning you can now carry v4 over v6 AF in the latest public release.

Effectively though – it is now possible to remove V4 entirely from P routers without losing any functionality – providing you engineer it right.  So we’re making progress 😊

Figured that might interest some

Thanks

Andrew

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.afnog.org/pipermail/afnog/attachments/20171002/cea8308f/attachment-0001.html>


More information about the afnog mailing list