[j-nsp] BGP Outbound Policy

[Pedro Roque Marques]

sshang2 at yahoo.com (Shixiong Shang) writes:

> Hi, folks:
> 
> I would like to compress BGP updates from Juniper M40s
> to other BGP routers. What I want to achieve is, Don't
> advertise the route learned from AS 65300 back to the
> routers in AS65300. By doing this, I think we can
> minimize the bandwith consumed by BGP and also improve
> the convergence time.
> 
> So what I did on Juniper was adding a policy statement
> as follow (the first a couple terms are to
> redistribute local, static and OSPF rotues into BGP):
> 

one caveat: the 'to neighbor x' clause will be ignored. BGP export
policy is applied once per group, thus one cannot specify different
policies on a per neighbor basis.

Thus you can apply this policy to your ebgp neighbors in 65300 but not
to neighbors in other ASes.

I would say that as far as being as optimization, it would only be
worthwhile if AS65300 advertises a significant number of active
routes to you. 
> 
> Neighbor routers 192.168.223.104 and 192.168.223.105
> are in AS 65300. My question is, if I apply this
> policy with term 50 and 60 to "protocol bgp export",
> will Juniper stop advertising the routes originated
> from 65300?

Yes.

> By reading the Juniper document, I believe
> the export policy only works between RIB and BGP
> routing table, not between BGP neighbors. Am I right?

Export policy can be used to filter/manipulate attributes for any BGP
advertisement. Not all policy qualifiers are valid in export policies.

  Pedro.