Why is it necessary to configure BOTH OSPF and BGP

dppagc

My WRONG understanding:

Equipment configured with OSPF will communicate with one another.
eBGP ensures communication of different routing protocols in other AS.

Since OSPF is configured for equipment within the same AS and they can already talk, why is ibgp necessary?
Is it another layer of protection?

I know that ibgp prevents routing loop. But it doesn't answer my question of why it "overlaps" with OSPF.

Sorry for my dumb question

_Gonzalo_

It´s not a dumb question.

The facts are:

BGP is necessary to communicate with other internet nodes.
iBGP is used to share BGP information with your own BGP peers. This way, when there are two different ISPs (which should happen often), routing information can be obtained from two different sources and for synchronism purposes has to be compared. How would that be done without an internal sub protocol, if eBGP routers are not connected?

And now, why not use OSPF for that?
The ammount of routes that BGP handles can be overwhelming for any other routing protocol. That´s why EGP and IGP exist, to cover different needs. As there is no need for all routers to know all routes (think about default static routes) so there is no need to saturate your OSPF protocol with all BGP routes. It is highly advised as per best practices NOT to redistribute from BGP into any IGP.

dppagc

Hi gonzalo,
just another question. Is ibgp only applied to the ASBRs and then ebgp is applied to the peering upstream?

rjon17469

dppagc wrote: »

Hi gonzalo,
just another question. Is ibgp only applied to the ASBRs and then ebgp is applied to the peering upstream?

iBGP is only needed to communicate BGP-specific information regarding routes between eBGP peers. You can have multiple eBGP peers with no iBGP connectivity and still have them exchange routing information through an IGP, but you won't have the full capabilities of BGP available this way. As Gonzalo mentioned, if you don't have multiple eBGP routers, you really have no use for iBGP.

The OSPF ASBR is just the point where external routes are injected into the OSPF routing domain. It could be anywhere along the BGP path, including on the device which has the eBGP peering, so that in a single router, single ISP environment no iBGP is needed.

ccie14023

It's actually not a dumb question, but a perfectly reasonable one. I remember being confused by this myself when I first started studying networking. There are some who think that there shouldn't actually be separate external and internal routing protocols, such as the eminent computer scientist Radia Perlman, who invented spanning-tree. If you are new to these concepts I recommend watching her keynote at the 2013 Linux conference in Australia. Although ostensibly about TRILL, she expounds on why networking can be so darn confusing. I also recommend her book, Interconnections.

In addition to the other responses, perhaps this will help. Think of BGP as an "edge" protocol that runs on the perimeter of an AS. It learns prefixes from other ASs using eBGP. These "edge" devices need to communicate directly with the other edge devices to share what they've learned. This is where we use iBGP. OSPF is more of an internal or core protocol. Now there are some tricks to how BGP interacts with the IGP (such as OSPF), but if you are just getting into this I don't want to overly complicate things. And yes, BGP can be used in non-edge scenarios too, but always start with the simplest usage and work your way to more complicated usages from there.

dppagc

By the way, when you say edge routers you are referring to asbr. Am I right to say that?
What other edge routers are there?

fredrikjj

ASBR is OSPF specific terminology for a router that redistributes an external prefix into to the OSPF domain. You can have a router that's at the edge of the network running BGP with another autonomous system without it being an OSPF ASBR.

networker050184

In most scenarios BGP routes are going to be recursive. Especially in a service provider environment. Next hops are going to be loopbacks. You want to take advantage of all of your IGP paths for fast failover to those next hops. You also want your label switched paths to set up correctly to loopbacks.

dppagc

fredrikjj wrote: »

ASBR is OSPF specific terminology for a router that redistributes an external prefix into to the OSPF domain. You can have a router that's at the edge of the network running BGP with another autonomous system without it being an OSPF ASBR.

Okay. Cam you an example of something that is not ASBR? Maybe I should use a generic term like border router.

dppagc

networker050184 wrote: »

In most scenarios BGP routes are going to be recursive. Especially in a service provider environment. Next hops are going to be loopbacks. You want to take advantage of all of your IGP paths for fast failover to those next hops. You also want your label switched paths to set up correctly to loopbacks.

Hi,

what do you mean by recursive?

networker050184

A recursive route lookup is when a route references another route as a next hop. Usually it will be the loopback of another router for ibgp routes known via igp. Another example would be a static route