[c-nsp] BGP peer/customer routes

[Pete Templin]

On 5/31/2011 10:50 AM, Tim Franklin wrote:
>> My standard practice has always been to apply a high local preference
>> on customer-announced routes, medium local pref on peer-announced
>> routes, and low (but still higher than the system default of 100)
>> local pref on upstream-announced routes.  The logic behind this is:
>> "I'd rather get paid for traffic than exchange traffic for free than
>> pay for traffic."
> Which is good design, but still falls apart if, as per the OP, your customer is deaggregating towards their other provider.  Longest-match happends before you even get into the BGP best-path algorithm, local-pref doesn't get a look-in.
In my particularly-humble opinion, it performs exactly as designed:

A customer who doesn't do anything special gets their traffic through 
their link "to me".
A customer who intentionally does "something" gets what they've programmed.
A customer who asks can apply communities to override our default BGP 
behaviors.

I'm only trying to change the default behavior of BGP, so if the 
customer gets creative with their announcements, they can do what they wish.

pt