IGP advantages with BGP

[hassan_syed6]

I have got a very basic and stupid question. It's just not getting through my head no matter how many document's i study on it.

My question is why would you need an IGP let's say OSPF in a network, where already BGP with "redistribute connected" and all the network statement's which needed to be advertised are configured. We also then redistribute the ospf into bgp. Mind you there is only one network statement under ospf.

Can someone direct me to a document which would give me specific information as to what are the advantages of using IGP with BGP. Even when BGP's are peering within the same domain, but there are couple of neighbor commands under bgp which are connected with the remote-as ( EBGP).

Hopefully, I am clear as to what is confusing me in my above explanation

[mheusinger · Accepted solution]

Hi,

First, for BGP session start you need IP connectivity, which is given through the IGP.

BGP sessions are setup through TCP port 179. In order to send an IP packet towards the configured neighbor IP you need a routing table entry for the neighbor IP.

For eBGP neighbors usually the directly connected interface IP is used, so the entry is in the IP routing table.

For iBGP usually loopback interface IPs are used - also remember the iBGP full mesh requirement! The loopback IP is not known outside the router unless announced through an IGP (or static routing is configured for that matter).

As long as BGP sessions do not start you can not learn about networks in your scenario. So the result would be no connectivity.

Second, BGP does not know of interface bandwidths . So choosing the best path internally between iBGP speakers is also a duty of the IGP. In addition your BGP should not be affected by network line outages internal to your AS. It is the job of your IGP to quickly route around those failed lines.

Hope this helps! Please rate all posts.

Regards, Martin