OSPF with BGP

[HamSolo]

Hi everyone, I'm a network engineer at a WISP, so I know a few things about networking (Although not at like CCIE level). Our WISP uses OSPF, with BGP just for IX peering. OSPF is starting to show it's limitations due to the size of our network. (We have 2,127 routes, and while it's ok now, trying to plan on our network doubling or tripling in size) Our network is currently a single area, and I know you can split it up into multiple areas, however what I'm looking to get more information on, is it's been proposed that we run BGP, with the variable "Next Hop" set to "Self" along with OSPF.  Does anyone have any ideas as I'm struggling to see how this would help, what it would do, etc... Any help explaining this would be greatly appreciated. Thanks!

P.S Using MikroTik gear.

[icecream-guy]

sounds like route summarization is needed on your network.  If those are all internal routes. Route redistribution can get tricky.
maybe a simple diagram would help.

[HamSolo]

sounds like route summarization is needed on your network.  If those are all internal routes. Route redistribution can get tricky.
maybe a simple diagram would help.

Thanks, we already do route summarization. I know one solution is to create multiple OSPF areas, but adding BGP so BGP is used for the customer routes, with next-hop self, and OSPF still used for loopbacks and management IP's was suggested as a possible solution as well.

[deanwebb]

And me the security guy just wants static routes everywhere so that there's no tomfoolery with the routing protocols...

Step one is to figure out what networks are in your backbone - those go to Area 0 (Backbone)

Standard areas will be ones that handle traffic between the backbone and areas further along. I consider those to be regional hubs, major office areas, campuses, things like that.

Stubby areas are ones are network ranges that will mostly route to each other and all other traffic goes out the default route, so they don't need a lot of routing info. This would be for things like small sales offices, retail outlets, other tiny networks that depend on resources mostly available outside their range.

So if it's something like a loopback or management IP, I'd consider putting them in regionally-aligned areas, where they'd all share the same basic routing information.

[HamSolo]

And me the security guy just wants static routes everywhere so that there's no tomfoolery with the routing protocols...

Step one is to figure out what networks are in your backbone - those go to Area 0 (Backbone)

Standard areas will be ones that handle traffic between the backbone and areas further along. I consider those to be regional hubs, major office areas, campuses, things like that.

Stubby areas are ones are network ranges that will mostly route to each other and all other traffic goes out the default route, so they don't need a lot of routing info. This would be for things like small sales offices, retail outlets, other tiny networks that depend on resources mostly available outside their range.

So if it's something like a loopback or management IP, I'd consider putting them in regionally-aligned areas, where they'd all share the same basic routing information.

Unfortunately static routes aren't much of an option, running a service provider network with currently over 150 POP's, over 5k customers, etc... static routes wouldn't scale well, lol.  Stub areas for OSPF wouldn't work that well either, while great if you're on a campus or enterprise, like one building could be one stub area, another building would be another, it's not a good way to build an ISP network, that's why we're looking at running OSPF for our loopbacks, and ibgp everything else, and was hopeing to get more information on that, if anyone has done it, pro's, cons, etc...

[deanwebb]

When I was at $GLOBAL_MEGACORP, with about 250 sites, we didn't have OSPF specifically for the loopbacks, etc.. They were all part of the general OSPF scheme for the region, with BGP running to sew up all the regions and keep them connected. That's about all I know, there. I'm a security guy.

Give a few days, and some of the other R&S guys might pipe up with an answer.

[icecream-guy]

for backbone routing IS-IS is the way to go, old protocol, but does a might fine job sharing routes very quickly.

[HamSolo]

for backbone routing IS-IS is the way to go, old protocol, but does a might fine job sharing routes very quickly.

Thanks, unfortunately the routing platform that we primarily use (MikroTik) doesn't support IS-IS so have to come up with another solution.

[Otanx]

Have you considered just doing BGP for everything? Give every router their own ASN from the private ranges, and then let them all peer with their neighbors. Using different ASNs for each router gets rid of the full mesh issue with iBGP. However, the config does get weird on any router that is peering outside of your network that needs to advertise with your real ASN and not the private one. I don't know if MikroTik can do that or not. I know Cisco can, but I don't remember what the feature was called.

-Otanx