OSPF Non-Broadcast Network Type over Frame-Relay

Hello Fazail

When you have a non-broadcast network type, in order to configure the OSPF neighbors manually, it is necessary to configure the neighbor command on only one of the two OSPF routers. Once this is done, the hub router will send out OSPF hello packets to the spokes to establish an adjacency. The spoke routers will respond to this and will create the neighbor adjacency.

Just to be sure, I’ve labbed this up myself as well, and after a few seconds, the adjacency does come up. You can see that this is also the case for EIGRP, as stated in this post:

I hope this has been helpful!

Laz

1 Like

Thank you so much Laz.

1 Like

Point to multi point network type can establish the neighborship by using the multicast address…Right ??.

@lagapidis you said that the point to multipoint network types is a collection of point to point thats right…but you said that there can a different subnet of each point to point connection …How ??

Frame relay and ATM and DMVPN,GRE , Over these overlay VPN we cant communicate between the OSPF neighbors by using multicast address 224.0.0.5…Right ??

Hello Narad

Yes, I mention that in this post. The statement was made not in relation to the subnets being used, but in relation to the fact that there are no DR/BDR elections. I have updated the text of the post for clarification.

Frame relay and ATM don’t support multicast natively, however, they can be made to support it in a pseudomulticast manner. As Rene mentioned in this post for Frame Relay, it is an NBMA technology, but with the correct frame-relay maps, you can “emulate” broadcast traffic.

DMVPN and GRE however are technologies that natively support multicast, so OSPF is able to operate over these networks. You can see in the lesson describing DMVPN Phase 1 with OSPF that the broadcast network type works well with DMVPN, as is the case with Phase 2 and Phase 3. DMVPN is based on GRE technology, so both similarly support multicast, thus, they support OSPF neighbors using the multicast address for communication.

I hope this has been helpful!

Laz

1 Like

Hello everyone!
Useful lesson!
I am a bit confused on using Non-Broadcast Network Type in a not fully meshed topology like this case.
Looking at the routing table of Spoke1, the next hop for the networks advertised by Spoke2 is Spoke2 itself and not the Hub.
Is there a way to forse the next hop to the Hub, like the point-to-multipoint non-broadcast network does automatically?
thanks

Hello Giacomo

The purpose of the non-broadcast network type is to make OSPF believe that you are running it on a multi-access network such as Ethernet. It is used to simulate such a network so that communication can indeed take place between spokes directly. The hub router has learned about the networks advertised by Spoke 2 and advertises these networks to Spoke 1 with a next-hop IP address belonging to Spoke 2. That way, communication takes place directly between the two spokes, even though they are not OSPF neighbors.

By definition, this is what the non-broadcast network type achieves in such a topology. You could force the hub to advertise itself as the next hop by using PBR, or route maps, or even static routing.

The question however is, why do you want to force the next-hop IP to the hub? If you want to achieve something specific, let us know, and we can help you find the best solution to achieve it.

I hope this has been helpful!

Laz

Question, we already used the neighbor statement here , so communication should be unicast, correct?, then whey we need this “Make sure you have a frame-relay map statement with the broadcast keyword or you won’t be able to send multicast on your frame-relay network” … thanks !

Hi Everyone, thanks for the explanation. Can you give us an example where prefered to choose one of each Types?

Hello Fran

  • Broadcast network type is the default for any Ethernet topology.
  • Non-broadcast network type would be used in a Layer 2 topology that doesn’t allow multicast and broadcast, such as a frame relay network. Here you must manually configure neighbors, otherwise no adjacencies would occur.
  • Point to point network type is used whenever you have a point to point topology between two OSPF routers, and you want to restrict the possibility of another router participating on the same segment. This can be configured regardless of the underlying technology (even over Ethernet).
  • Point to Multipoint network type acts as many point to point networks. This again is useful with frame relay networks that support “pseudo-broadcast”.
  • Point to multipoint non broadcast network type acts the same as point to multipoint, but can be employed in frame relay networks where pseudo-broadcast is not supported.

As you can see, some of these network types are specifically designed to allow OSPF to run over non-broadcast Layer 2 technologies such as Frame Relay. As these technologies are being phased out, OSPF network types are being used less and less. There is an exception, however, and that has to do with DMVPN. You may choose to use a particular network type with DMVPN depending upon your topology. Take a look at these lessons which show how the various network types are applied to different DMVPN Phases:

I hope this has been helpful!

Laz

1 Like

Hello,

Thank you so much for exciting Lessons.

This lesson uses a Frame Relay environment to validate OSPF non-broadcasts, but will Frame Relay appear within the CCIE Enterprise Infrastructure test scope?
Also, in my home verification environment, I am using EVE-NG’s Community Edition, so I cannot reproduce the unsupported Frame Relay environment … Which emulator is recommended for reproduction?

Regards

Hello Yoichi

Frame relay as a technology is no longer a topic on any Cisco exams. However, OSPF network types are. The most convenient way to simulate network types such as non-broadcast and point to point are to use frame relay as the underlying technology. Take a look at this NetworkLessons note on frame relay for more information.

If you want to reproduce a frame relay environment, you can do so using GNS3 or VIRL/CML.

I hope this has been helpful!

Laz

Hi Laz

Thank you so much always.
Today, I bought CML! (It was discount Price.25% off.)
I try to reproduce a frame-relay environment on CML!!

Yoichi

1 Like

Hi
If i have network with many routers (around 60 router) and all connected through isp in one broadcast domain (one vlan and one subnet)
Sometime i cause flapping bcz of sfp flooding and sfp recalculation, all interface with default network type which is broadcast
How can i optimize and tuning my network type to stay within one broadcast domain and one subnet?
My design is hub and spokes but not running dmvpn or anything Advance.
@ReneMolenaar @lagapidis

Which network type better for me?

Hello Ali

First of all, having 60 OSPF routers in a single broadcast domain will cause you problems. And if I understand correctly, you have a hub and spoke topology, so all of the remote routers are listening to broadcast traffic on this single subnet over their WAN connections. So you’re using up bandwidth unnecessarily, and you are also causing the routers to look at broadcasts that do not belong to them, so they continually discard them.

The best way to resolve this issue is not to choose another OSPF network type, but to change your topology so that each spoke belongs to a different subnet. That way, you can employ routing more efficiently and send traffic only to the spoke for which it is intended. Using either FlexVPN or DMVPN would be best.

I don’t believe that any choice of network type will resolve your flapping routes, your SFP flooding and the broadcasts being sent.

I hope this has been helpful!

Laz

@lagapidis
i can’t use for each spoke subnet as long as i got one vlan from my isp, in this case what can i do? I fixed this issue by another way, i did QinQ and used inner vlan in each site and ospf over this inner vlan, but was hard way, is there any good solid solution rather than this one?
What if i used another network type like p2mp nb?

Another Q, as i know just NonBroadcast and P2MP is open standard and all other network types are cisco proprietary, but i saw some of these in mikrotik router, could you clarify this point plz?

Hello Ali

The solution of using QinQ is clever, and indeed does resolve the issue, but I agree that it is somewhat cumbersome.

Even if you have a single subnet from your ISP, one solution that may be worth looking at is to configure DMVPN. You can configure DMVPN Phase 3 with OSPF. This will allow you to have spoke-to-spoke communication where NHRP is used to share routes in conjunction with OSPF. It is much better than having a single subnet with 60 OSPF routers. It resolves the issue of too much broadcast traffic and missed OSPF updates and flapping routes.

Now in the DMVPN OSPF configuration, you are able to change the network types. This should not be confused with the network types you would have changed if DMVPN was not used. These network types are configured within the tunnel interfaces. For a large number of spokes, I believe the point to multipoint network type would be the most efficient. In the lesson that describes it here you can see in detail how it works.

That would be my suggestion, but I would advise you to first reproduce it with multiple spokes in GNS3 or CML or some other emulator, to make sure you have the behavior you are looking for.

Let us know how you get along!

I hope this has been helpful!

Laz

Thx for your reply
What about second Part (second Q)?
Please advice

And also regarding first part
What if i have multi vendor devices? Some of them cisco and other Mikrotik which doesn’t support dmvpn!

Hello Ali

Yes, according to RFC2328, OSPF defines three network types: Point to Point, Broadcast, and Non-Broadcast. Cisco has added the Point to Multipoint and the Point to Multipoint Non Broadcast networks, as described in the lessons.

Now as it happens with many technologies, Cisco has been the first to implement them, and other vendors choose to also implement them in order to maintain as much compatibility as possible. For example, other vendors support CDP, as well as EIGRP, both of which are Cisco technologies. Mikrotik as well as many other vendors, seem to have chosen to support these other non-standard OSPF network types in order to maintain compatibility with Cisco.

Indeed Microtik doesn’t support DMVPN. I don’t have a specific solution for you, however, there are several workarounds that can be used. You can find a lot of documentation that describes how to create a DMVPN-like network using Microtik equipment.

In general, when you have such a large broadcast domain, it would be best to redesign the network from scratch. In other words, find a way to have your ISP provide you with some form of subnetting/routing so that you can avoid broadcasts over your WAN. Alternatively, could you ask your ISP to use IPv6 addressing which is able to manage large broadcast domains much better?

I hope this has been helpful!

Laz

1 Like

Hi Rene,
I have a question. Why you specified statically neighbor on HUB only? How does it work that neighbor adjacency is established. Additionally when I set neighbor command on spokes and i delated it from HUB i couldnt form neighborship.
Best regards,
Marcin

Hello Marcin

When statically configuring OSPF neighbors, it is only necessary to issue the neighbor command on one of the two OSPF routers. It will send a hello to the other informing it of its intentions and its IP address. The other will respond and create a neighbor adjacency. However, to be sure, it is always best practice to configure the neighbor command on both routers rather than just on one.

Hmm, I’m not sure why this is, but I suspect it has to do with the priority 0 that has been configured. The spokes are set not to become DRs. This is just a hunch, however, you can try removing that command to see if it affects the results of your experiment.

I hope this has been helpful!

Laz

1 Like