[c-nsp] New IPv6 BGP peer on a pure IPv4 network

[Gert Doering]

Hi,

On Tue, Dec 02, 2008 at 02:17:59PM +0200, Ziv Leyes wrote:
> Well, when you say it,  it sounds very simple, the problem is I 
> don't really know the subneting stuff for IPv6, for example.

Well, it's like CIDR in IPv4 - you put aside a number of bits for the
"network" part, and the rest is "host".

In IPv6, it's actually easier, as you don't need to do much thinking - just
make every network a /64, and all loopbacks a /128.  

(This is in line with IETF and RIPE policies, even if it might seem like
"waste" - but there are enough addresses - and even if some folks do it 
differently for one or the other reason.  No need to complicate matters
here).

From RIPE, you will get a /32.

Every customer gets a /48 (65000 of those inside /32).

Every of your POPs gets a /48.

Inside the /48, there's 65000 /64s.  So use them freely, do not worry about
waste here.

[This is simplified, for a "really large" provider, the math is more 
complicated, of course]

> We don't use any of the IGP you've mentioned in our IPv4 setup, we only have some iBGP  peers between our routers.
> Do we HAVE to use OSPF, ISIS or EIGRP or we can still use only iBGP on IPv6 when needed? 

If you don't run any IPv4 IGP today, you can use the same setup for IPv6
(static routes + iBGP + eBGP).

> Remember I'm not moving the whole network to v6, only additional to the current v4.

Just add IPv6 to the existing config.  It will run in parrallel.


> Let's say I have a circuit connection to one uplink provider and I have 
> a BGP peering between both sides on IPv4 /32 addresses, you suggest 
> that I ask the provider to move to IPv6, what happens then to the whole 
> IPv4 addresses? 

Add IPv6, but do not remove IPv4.  Run both in parallel.

This is, for example, how our uplink interface to C&W looks like:

interface Port-channel150
 description GE-Link zu C&W 1273
 ip address 62.xx.xx.162 255.255.255.252
 ip access-group 110 in
 ip flow ingress
 ipv6 address 2001:yyy:yy:yy::2/64
 ipv6 traffic-filter ext-in in
end

the IPv4 part is easy to understand - and the IPv6 part essentially does
the same thing, just with funny numbers, and a named IPv6 ACL.

> What happens to other devices that know this IPv4 address and won't know 
> the new IPv6 address? 

They continue to use IPv4.

> What about the loopback IPs that are used for other peering and stuff?

interface Loopback0
 description Loopback, fuer iBGP
 ip address 193.xx.xx.14 255.255.255.255
 ipv6 address 2001:608:0:zzz::1030/128
 ipv6 ospf 42 area 0
end

- again: add v6, leave the v4 part alone.

> Is there a way to transfer IPv4 traffic through an IPv6 BGP and vice versa?

Yep.  Tunneling - a GRE tunnel (for example) which carries IPv4 inside and
IPv6 outside, or vice versa.

But for a 3-router network it's much easier to enable both protocols in
parallel.  Tunnels are a workaround for underlying infrastructure constraints.


(NB: there's a number of Cisco Press books about IPv6.  I have not yet
looked at one of them, so I can't say anything about the contents - but I've
met the authors and they know what they are talking about - so chances are
quite high that the books are worth looking at)

gert
-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert at greenie.muc.de
fax: +49-89-35655025                        gert at net.informatik.tu-muenchen.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 304 bytes
Desc: not available
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20081202/a5ccb06d/attachment.bin>