From: Roman Rodichev (rodic000@xxxxxxxxxxx)
Date: Sun Jun 03 2001 - 02:03:51 GMT-3
This is a very good questions, and I think it has been discussed once here.
You really have to look at this question from a real-world perspective. If
you do redistribution of IGP routes into BGP (to be sent to other autonomous
systems) only in one point of the network, then you are actually relying on
IBGP to transmit those internal routes to other IBGP peers connected to
other autonomous systems.
Let's say you have an autonomous system consisting of over 100 routers
running OSPF. On the edges you have 10 BGP routers connecting you to 10
customers in different autonomous systems. BGP routers run IBGP between
them. Now let's say you decide to redistribute all of your ospf routes only
into one BGP router A, and that BGP router A will send the routes to IBGP
neighbors for remote delivery to other autonomous systems. What happens if:
- That BGP router A is taken offline for maintenance?
- Network breaks somewhere around that router and all IBGP sessions are
dropped?
What will happen is all customers connected to BGP routers other than A will
- Loose routes of the customers connected to the router A (which is not a
big deal)
- Loose all of your internal routes. Those customers will have routes to
your other customers (IBGP is still in place), but will not know of your own
routes.
Bad idea! Bad! That's why you would prefer to redistribute your ospf routes
on all BGP routers. And if you want to have some fun, you can actually get
rid of redistribution completely and implement MPLS. On the edges your BGP
routers would run EBGP and IBGP. And internal OSPF domain would know only
about internal routes that will provide paths for IBGP next-hops. The
assumption is that your network is a core network providing connectivity
between the customers, and that you don't need to see your customers routes
on your internal routers.
The BGP routers on the edge instead of assigning a label to each IBGP route
that came from other autonomous systems, it will assign the same label that
was assigned to the route of IBGP next-hop. While traversing your core
network label swapping will occur and the packet will appear on the other
edge. From that point it will be Layer3 routed using EBGP routing table.
Sorry... I got carried away. I hope you can understand now the importance of
redistributing OSPF->BGP on multiple routers.
Roman
>From: "Mario Rivas" <mrivas@desca.com>
>Reply-To: "Mario Rivas" <mrivas@desca.com>
>To: <ccielab@groupstudy.com>
>Subject: CcieBootCamp Lab 8
>Date: Sun, 3 Jun 2001 00:15:21 -0400
>
>Hi Guys,
>
>I need help from somebody about cciebootcamp lab 8.
>
>Why in redistribution from ospf to bgp have to do in r6 and r1; anybody
>has probe in only one.
>
>Why in redistribution from osfp to bgp route map is span in both routers
>
>Thanks
>
>Mario Rivas
>**Please read:http://www.groupstudy.com/list/posting.html
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:31:17 GMT-3