From: Howard C. Berkowitz (hcb@xxxxxxxxxxxx)
Date: Fri Apr 26 2002 - 22:23:00 GMT-3
At 11:44 PM +0000 4/26/02, kym blair wrote:
>I know the mechanics of how to set up both route-reflectors and
>confederations, but I'm not sure when to use one or the other, and
>where to place the route-reflector or various confederations.
>
>How would you solve the following example:
>
>AS2-----AS1--------------AS1-----AS2
>
>.........R1----R3---R4---R5.........
>
>All four BGP routers are in AS1. All four (plus R2) are known via
>your IGPs. R1 and R5 are both connected to different routers in AS2
>(outside your control). You'll want to manipulate weight, local
>preference, community settings, etc via R1 and R5.
Well, it's hard to decide with just this information. First, let me
make some general observations. As you know, both reflectors and
confederations were means of making iBGP more scalable.
I'd start by drawing it a little differently.
AS2 POP 1 AS2 POP 2
| |
| | eBGP
| iBGP |
R1...............................R5
| |
+------R3--------------R4-------+
IGP
It's hard to generalize, but reflectors probably are easier to
configure* and impose less overhead. Confederations give more
control. In my experience, I've tended to use reflectors in
relatively homogeneous service provider networks, but configurations
in the backbone-of-backbones of an enterprise.
* Things can get quite complicated when you use hierarchies of reflectors
and/or do a fair bit of IGP-BGP redistribution. There is an Internet Draft
in the IDR working group that discusses the "BGP Persistent Oscillation
Condition" that comes from things like this. For example, you have to
work your redistribution so you have something OSPF-like. Where OSPF alway
s
prefers intra-area routes, you want POPs to always prefer intra-POP/intra-
cluster routes.
Confederations can also be useful when you are merging or divesting
ISPs, and the ISPs have different policies.
>
>Would you set up a route-reflector with 3 clients, or would you set
>up confederations? Where would you put the route-reflector(s) or
>how would you set up confederations for best results? Also, very
>important: which router(s) would you use to redistribute between BGP
>and IGP?
With 4 or 5 routers, I wouldn't use either. I'd just go full mesh.
From what I see, I'm not sure why you would need to redistribute
between them at all, except, of course, for the strange purposes of
the CCIE lab. In general, if you have to, it's probably best to do
it at your eBGP speaking routers.
>
>I've read Halabi, Doyle Vol 2, and Parkhurst and understand the
>mechanics, but am not good at deciding where to put things. What
>section of these books (or others) should I reread to get a better
>understanding of bgp network design?
I have one book out and one coming out in June that deal with
provider/BGP network design rather than configuration, both from
Wiley. The one out now is the "WAN Survival Guide," which is more
focused on the enterprise side of the enterprise-ISP interface. The
new one, called "Building Service Provider Networks," complements it.
Hopefully, I will have an anonymous FTP server with some of my
presentations on this up next week, but, in the meantime, you can
navigate around www.nanog.org and look in the meeting archives. At
the very least, you'll find tutorials by Paul Ferguson, Avi Friedman,
and myself.
In a few weeks, I should have a virtual rack available with external
generation of complex BGP routes, more than you can do in a
six-router pod. The scenarios for this will be open source (with
some licensing caveats), although the traffic simulator will not.
>
>Thanks, Kym
-- "What Problem are you trying to solve?" ***send Cisco questions to the list, so all can benefit -- not directly to me*** ******************************************************************************* * Howard C. Berkowitz hcb@gettcomm.com Chief Technology Officer, GettLab/Gett Communications http://www.gettlabs.com Technical Director, CertificationZone.com http://www.certificationzone.com "retired" Certified Cisco Systems Instructor (CID) #93005
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:58:20 GMT-3