[c-nsp] BGP peer/customer routes

[Mark Tinka]

On Wednesday, June 01, 2011 06:31:42 AM Peter Rathlev wrote:

> I'm seeing this from a customer perspective: Why on earth
> should you not respect the more specific routes via the
> peering link?
> 
> What if I have a primary connection from AS11 and buy a
> backup connection (much lower bandwidth) from you, but
> another of your customers is the new Youtube? If you
> insist on sending traffic from them down the backup pipe
> I bought from you it wouldn't work. I'd have to find
> another ISP to buy transit from, and sending it upstream
> is at least more expensive than towards a peer, right?

We see this all the time, and our general policy is not to 
peer with customers. However, if our customer is buying 
transit from a competitor that's peering with us at the 
exchange point, we would generally pass that traffic over 
since the alternative is not preferred (via our own tran$it 
links).

It's a very hard problem to police, as customer routing 
policies can be as dynamic as your own, if not more.

So as long as we're not peering with our customers, we can 
bend a little. If there are things we can control/fix 
scalably, we do that. Else, it's better than the 
alternative, because if a customer doesn't want to announce 
more specifics to you, they don't want to announce more 
specifics to you.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20110601/f5899ddf/attachment.pgp>