BGP as IGP

[michael.luo]

We are building an enterprise network with MPLS VPN (service provider MPLS).  We originally planed to use OSPF on internal routers and BGP on CE routers.

Someone said why don't you use BGP all around?  So you run only one routing protocol instead of two?

I don't get it.  If we ran BGP, we still need IGP as the "under layer", otherwise we'll have to use static routes all over (which is crazy).  So how could I "save one routing protocol"?

Also, what's the advantage of BGP in internal network vs. OSPF?  Thanks!

[Philip D'Ath]

I love EIGRP.  It is so easy to manage, and doesn't have so many horrible design restrictions as compared to OSPF.  It is my first choice for interior routing protocols.

I like iBGP.  It has so many knobs and dials you can turn to adjust almost anything you can think of with regards to routing policy.  eBGP is a universal standard for interfacing between networks, and it is easy to carry all sorts of different metrics from from AS to another.  You can then choose to move these directly into your iBGP if needed.

I have frequently used iBGP and eBGP together.  I often use iBGP when I know a network is likely to have complex interior routing policy requirements.  xBGP also plays nicely with other routing technologies like Pfr (performance routing).

I hate OSPF with a passion.  I would only use it if you want to remain vendor neutral.  However it is still very popular for what you describe (interior routing protocol for MPLS).  OSPF is usually one of my last choices for an interior routing protocol.  But it is good if you are a sadist.

[Mark Malone]

BGP is not an IGP it operates at layer 4 of the OSI on top of a layer 3 IGP you need the IGP as you said , it was built as a transport layer protocol not a routing protocol there is ISPs out there that use OSPF right through there ISP networks without BGP but there few and far between

There would be no real advantage of using BGP in an internal network , you want an IGP that converges fast , BGP does not work on convergence speed so its not good idea to try and use it as an IGP

The reason most people use BGP is because most major ISP providers use it as there EGP so you don't have a lot of choice, there is methods though which allow you tunnel through their networks at layer2 which would allow you to avoid using there BGP and you would just connect through the ISP using your IGP making the BGP/MPLS transparent to you

BGP is very flexible/stable and that's just some of the reasons its used so widely globally

[michael.luo]

I don't see any advantage (using BGP in internal network) either.  Not sure why Cisco SE tried to sell this idea.

[hgade]

This would be similar setup like this : https://tools.ietf.org/html/rfc7938 : We can use BGP as point-to-point links where directly connected will be the next-hop value and IGP. Reason to use BGP instead of any other IGP is because we can hop by hop traffic engineering, use one routing protocol and extensible.

[Julio E. Moisa]

Hi Michael,

There are many ways to build a network although you could use BGP as an IGP (although it is not considered as a IGP protocol) you must consider the convergence times and in certain way complexity of this protocol. For example if you are going to use iBGP you should create a full mesh to avoid loops otherwise you will need to create route-reflectors or confederations. 

A network administrator should have a good BGP knowledge to avoid any inconvenience after any chance or implementations. Also you could use an IGP protocol to be the NLRI as you mentioned (underlayer) in order to have flexibility and quick high availability but you could overloading the device with 2 routing protocols. From my point of view I prefer to use EIGRP or OSPF internally and keep BGP on the Edge routers.

Hope it is useful

:-)