[c-nsp] Converting OSPF backbone to iBGP

[Mark Tinka]

On Tuesday 30 September 2008 23:56:29 Dan Armstrong wrote:

> Unfortunately our access network is highly distributed,
> consequently our address space is highly de-aggregated at
> the access layer.  We shove our Supernets into iBGP
> regionally and currently let our IGP sort out the lower
> layers... which is what I want to get away from.

In our experience, there are 2 types of routes at the edge:

a) the point-to-point address used to connect the customer
   (/30 for v4, /126 for v6).

b) the assignment given to the customer for further use on
   their LAN (/29 for v4 minimum and/or shorter, /48 for
   v6).

For a), we typically assign at least a /24 (v4) or /112 (v6) 
to an edge router and announce that via iBGP. This covers 
all /30's and /126's we'd use for individual customer 
point-to-point links.

For b), as in your case, these can be distributed, 
particularly if customers have to move around edge routers, 
e.g., to obtain different services, get connected to a 
higher capacity router, e.t.c., and expect to keep the same 
IP addresses whenever possible. We generally don't mind if 
these are disjoint or non-contiguous (BGP was built to 
scale with the growing number of route entries, internal or 
external). The most important thing is to have a definite 
way to track assignments across the network. Of course, 
wherever possible, aggregate.

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20081001/36357826/attachment.bin>