Re: [nsp] MPLS Traffic engineering

From: dhudson (dhudson@pilot.net)
Date: Tue Nov 07 2000 - 11:16:48 EST


Eric Osborne wrote:
>
>
> > What about iBGP? I know somebody who succefully running MPLS network using
> > only iBGP as IGP. Sure, it requires careful network design and addressing
> > planning, but it pays off by actually working. ;) Please correct me if I'm
> > wrong somewhere.
> >
>
> MPLS-TE (Traffic Engineering) needs a link-state IGP.
> MPLS forwarding based on LDP/TDP (whether or not you run VPNV4BGP on
> top of that) doesn't care what your IGP is, although personally I have
> my objections to using iBGP as an IGP...:)
>
> eric

i think running iBGP as an IGP has some major drawbacks.

1. it is meant to tell other iBGP speakers about the best pathway
OUT of the AS not really optimal for internal navigation.

2. if you arent fully meshed, then you are running in a degraded fashion.

3. it converges on an order of magnitude slower than any of the IGPs.

4. you violate the first cardinal tenant of when to use any BGP
if you have any stub nets with a single entry/exit point.

5. because the iBGP speakers must communicate with eBGP, it creates the
potential for allowing an external AS alter your routing policy.

6. if you are connected with an ISP and your internal policy is the same
as theres, BGP shouldnt even be running at all. statics and defaults
are cleaner.

7. why when you have OSPF...

piranha...lurk whether we are hungry or not...

-- 

--------------------------------------------------- my lord tzu, running away is the first martial tao archery sifu to sun tzu ---------------------------------------------------



This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:20 EDT