Using iBGP over OSPF within LAN

[PAHEE87]

Gentlemen,

We have a multi-vendor setup in our datacenter(Silverpeak for SDWAN and Cisco Nexus devices). What could be the reasons that we should deploy iBGP over OSPF within LAN in the data center?

Thanks

[pman]

Hi,

Using an IGP (like OSPF or ISIS) is recommended with iBGP.

Without the IGP, iBGP must neighbor on external-facing interfaces (physical interface),

with an IGP (like OSPF or ISIS), iBGP can establish between neighbors on loopback interfaces, and can have multiple paths to reach each neighbor.

that's because the loopback can be advertised from multiple physical interfaces.

[Richard Burts]

I do not understand the question from the original poster. I think I understand that there are several devices connected on a LAN. I assume that these devices are running EBGP with external peers. It would make sense to have BGP "with" OSPF. IBGP so that the BGP routers would communicate external routes with each other and OSPF to communicate internal routing information. But I do not understand BGP "over" OSPF.

[MHM Cisco World]

In Data Center ibgp over ospf "ospf underlay" meanly use for vxlan.

Keep in mind that BGP can carry MAC address over L3 network.

[paul driver]

Hello

As you mention sdwan/nexus and ibgp over ospf, it does suggest a possible vxlan deployment (spine/leaf) with the vteps edges (leafs) being ibgp route reflector clients towards their route reflectors (spine) with ospf being the the transit path for the vxlan

[PAHEE87]

Sorry Gentlemen. I should have worded more appropriately. But original question to you is "I know we can run BGP in data center within LAN. I would like to know a scenario where we deploy BGP than using OSPF on a multi-vendor setup in the LAN in the data center( Assume that we are using ebgp for WAN)". Sorry if this question is too vague. 

[Elliot Dierksen]

Like most things in a deployment, it depends on a bunch of factors. If you have a bunch of devices involved within the DC, that would be a lot of BGP peers since you have to do a full mesh. You could do route reflectors and/or confederations, but then the complexity goes way up. Bear in mind that BGP is built for stability, NOT rapid convergence. I would think you would want to rapid convergence within a DC which means some kind of IGP (OSPF, EIGRP, ISIS). Apologies if I am saying things you already know, but this brings in the differences between an IGP (interior gateway protocol for use between routers under common administrative control) versus EGP (exterior gateway protocol for use between routers under different administrative control).

[MHM Cisco World]

depend on SP and your DC requirement. 
depend on SP if SP config it side with BGP then you force to use BGP in your side. 
depend on your DC requirement 
I have DC-A and DC-B 
and I need to interconnect them, 
both DC use VXLAN then I must use eBGP to carry all L2 attribute between two DC and appear that both DC is L2 connect but in real then interconnect via L3 which will use as underlay.