Re: single point of redistribution

From: Howard C. Berkowitz (hcb@xxxxxxxxxxxx)
Date: Sun Aug 25 2002 - 17:24:04 GMT-3


   
At 11:19 AM -0700 8/25/02, ccie candidate wrote:
>guys ;
>
>when we have single point of redistribution ..is there any need for
>filters ?? and why or in which cases this is necessary
>
>thanks

Let's distinguish between two kinds of redistribution: hierarchical
and mutual.

In hierarchical redistribution, you generally have a more capable
classless IGP being fed by a less capable classful IGP, or BGP fed by
an IGP. At a simple level, you want to tell the more capable protocol
what you know, and take only default from it.

 From less capable to more capable: deny default, permit everything else
     Do summarize when appropriate.
 From more capable to less capable: permit default, deny everything else.

This is the way most reliable real-world networks are set up when
bringing in legacy routing, new acquisitions, etc. The principle
still holds if you have multiple points of attachment, although you
may want to try to get load-sharing across the defaults. Unless you
are running OSPF and can make the default a type 1 external, this
usually means you adjust the metrics so router 1 is closer to half
the routers and router 2 is closer to the remaining routers. If
either router goes down, everything switches automatically to the
other default router.

Mutual redistribution is where you always need filters.

I have had design cases where things were a blend of the two kinds of
redistribution. In one case, the networking research lab of a large
telecommunications corporation already had special, very-high-speed
connections to the Internet, faster than that of the corporate
Internet gateway. In this case, I generated the preferred default
(using a lower OSPF type 2 metric) from the division's Internet
gateway, and made the corporate default a backup. I advertised
summarized corporate routes into the division, and, for those that
should always be inside the enterprise, filtered them inbound at the
division gateway. By using OSPF type 2 metrics, we could receive the
same route from the corporate core and the division gateway, and
always prefer the corporate path unless it was down but another
corporate unit still had connectivity via the Internet.



This archive was generated by hypermail 2.1.4 : Sat Sep 07 2002 - 19:48:37 GMT-3