iBGP instead of OSPF?

[louis0001]

I've asked a few questions here about this setup and the more I speak to some very knowledgeable people on here, the more it gets me thinking.

Basically we have 2 core sites connected via a 1gb PtP link. We use OSPF to route between these sites using R1 & R2. Nothing fancy really, about 6 subnets per site advertised via OSPF to each other.

At each site, we have another router (R3 & R4) that each peers into an ISP provided MPLS network that doesn't have any internet breakout etc ie it's totally private.

Should I consider removing the OSPF between R1 & R2 and change to BGP to simplify things?

Current setup:

Remote sites > R3 (BGP) > R1 << PtP (OSPF) >> R2 < R4 (BGP) < Remote sites

Should I change to:

Remote sites > R3 (BGP) > R1 << PtP (BGP) >> R2 < R4 (BGP) < Remote sites

NB: The 50 remote sites (above) are the same and all of them have a route via BGP to R3 & R4

[balaji.bandi]

Personally I do not see any advantage here going to iBGP for this small network.

If any other reason or limitation you thinking to go iBGP ?

[louis0001]

There are 50 remote sites (BGP) and other networks hanging of the two core sites but nothing major. I basically want a bit of redundancy for the remote sites which are connected to both sites via BGP. So rather than use both OSPF between R1 & R2, I was thinking of simplifying it and taking OSPF out of the equation and purely use BGP between all sites.

That way, if the outside interface of R3 failed, a remote site could reach R1 via R4 & R2?

[balaji.bandi]

Sure, Agreed after your more explanation of setup.

you can create mesh network using iBGP, if you happy with that.

[louis0001]

Could I ask how I would achieve the above?

Would it be like:

1. Bring up BGP on R1 (eg. AS 65001) & R2 (eg. AS 65002) and make R1 & R2 neighbors

2. Remove OSPF from R1 & R2

2. Make R1 (AS 65001) & R3 (AS 65003) neighbors

3. Make R2 (AS 65002) & R4 (AS 65004) neighbors

Should I make R1 & R4 neighbors, R2 & R3 neighbors too so that each router has each other as a neighbor?

Basically, I want to simplify the routing between R1 - R4 by removing OSPF & static between R1 & R2 and just use BGP so that any changes are reflected throughout and AS 65010 - 60 can reach eg. R1 if R3 failed etc.

[balaji.bandi]

This topo is HLD, until we see all the configuration, i would be in comfort position to suggest.(this is only a suggestion, again we dont know the dependencies and applications running, how they configured).

But if the network is critical always suggest to PS Service from cisco partner, so they can asses your network and give you downtime or seam less change.

Make Sense ?

[louis0001]

This is just an idea I'm toying about at the moment and would be fully tested before being deployed. It's not much more complicated than what is shown and we're only trying to simplify it down and add a little redundancy. 

The basic question is "Am I better off using BGP for the internal routing rather than a mix of OSPF & BGP"

Am I better putting R1 & R2 into the same AS or just giving them separate private AS no's?

[balaji.bandi]

I always put our internal network own AS.

[louis0001]

So, in the above, would I:

1. R1 & R2 in same AS as neighbors

2. R1 - add R3 as neighbor?

3. R2 - add R4 as neighbor?