OSPF Point-Point vs Broadcast Network Type (Advantages)

[nwekechampion]

Hi all,

What would be a major advantage of using p2p network type over broadcast network type.

from what I can see, as far as convergence goes they have

1. the same hello and dead timers

2. They both send multicast packet regardless to 224.0.0.5 (for ospf neighs discovery)

I guess my point is, why would I want to use or choose p2p over bcast?

[M02@rt37]

Hello @nwekechampion,

P2P networks are inherently more scalable than broadcast networks. In a broadcast network, each message is sent to all nodes in the network, which can lead to increased network traffic and resource utilization as the network grows. In contrast, P2P networks distribute the load by allowing nodes to communicate directly with each other, reducing the burden on individual nodes and enabling better scalability.

P2P networks are designed to be more resilient to failures compared to broadcast networks. In a P2P network, if a node goes offline or fails, other nodes can still communicate with each other directly, maintaining network connectivity. This decentralized nature of P2P networks ensures that the failure of a single node has minimal impact on the overall network.

P2P networks can provide enhanced privacy and security compared to broadcast networks. In a broadcast network, all messages are visible to all nodes, potentially exposing sensitive information. In a P2P network, nodes can establish direct connections with trusted peers, allowing for encrypted communication and selective sharing of information, thereby enhancing privacy and security.

The choice between P2P and broadcast network types depends on the specific requirements and characteristics of the network. While P2P networks offer advantages in certain scenarios, broadcast networks have their own strengths, such as simplicity and ease of implementation.

[David Ruess]

I would argue these are more generic traits of P2P vs Broadcast network and not related to OSPF. I humbly ask you to consider the following:

---

M02@rt37 wrote:

Hello @nwekechampion,

P2P networks are inherently more scalable than broadcast networks. In a broadcast network, each message is sent to all nodes in the network, which can lead to increased network traffic and resource utilization as the network grows. In contrast, P2P networks distribute the load by allowing nodes to communicate directly with each other, reducing the burden on individual nodes and enabling better scalability.

Kinda. OSPF combats this by electing the DR/BDR on broadcast networks to prevent this exact scenario. So in an OSPF broadcast network each node does not send to all nodes in the broadcast domain as you mention. Its sends to a multicast address meant only for one device (DR) and that device distributes the LSU packets.  Also note that in OSPF whether its P2P or Broadcast all LSU packets are sent within the Area. So even in a P2P network LSUs are sent to all nodes regardless, its just passed from Router to Router.

P2P networks are designed to be more resilient to failures compared to broadcast networks. In a P2P network, if a node goes offline or fails, other nodes can still communicate with each other directly, maintaining network connectivity. This decentralized nature of P2P networks ensures that the failure of a single node has minimal impact on the overall network.

This is not quite correct. Design scenario: If you have 3 routers connected in a line with P2P network connections and the middle router fails....you have lost communication to all devices (becasue the end devices cannot communicate through the dead router). On a broadcast network if 1 router fails the other 2 can still communicate with eachother.

P2P networks can provide enhanced privacy and security compared to broadcast networks. In a broadcast network, all messages are visible to all nodes, potentially exposing sensitive information. In a P2P network, nodes can establish direct connections with trusted peers, allowing for encrypted communication and selective sharing of information, thereby enhancing privacy and security.
This is not correct. You can do encryption on a broadcast network. Its also built into the OSPFv3 protocol

The choice between P2P and broadcast network types depends on the specific requirements and characteristics of the network. While P2P networks offer advantages in certain scenarios, broadcast networks have their own strengths, such as simplicity and ease of implementation.

---

As @Joseph W. Doherty mentioned and you did in an earlier comment it reduces the LSA type by 1 and forgoes the DR/BDR election as its main benefits. The rest of the benefits come as an "it depends" kinda scenario, also mentioned by @Joseph W. Doherty  Hope this adds a bit of clarity for the OP as well.

-David

[nwekechampion]

Also @David Ruess  just figured.. that point-point on loopbacks install a network on rib as opposed to /32 by broadcast as well.

Not exactly sure why as I thought it'd be the reverse

[David Ruess]

In OSPF on Cisco devices the Loopback is its own network type of "Loopback". If you change it to P2P or broadcast (or any other network type other than Loopback) it will advertise the correct mask. If you keep it as the Loopback network type it will advertise it as a /32 regardless of the mask. P2P doesnt mean a /32 - if anything its a /30 or 31.

-David

[MHM Cisco World]

I dont see any different from my view, keep default is best.

[Joseph W. Doherty]

On a shared media, like across an Ethernet link, two OSPF routers establish full adjacency, noticeable faster, when configured at OSPF p2p.  Once adjacency is established, unless that particular adjacency is dropped, I suspect, dealing with other topology changes doesn't matter whether p2p or broadcast network type (for just the pair).

If you have the situation, with a core L3 switch with links to multiple distribution L3 switches, you have the choice of placing all those routers into one VLAN or making each physical link its own p2p.  In that case, which is better?

Well, that's an "it depends", for example, if you're supporting multicast.  With multicast between routers, all on a shared broadcast media, without Cisco's PIM snooping, you have a situation somewhat like multicast has on access networks without IGMP snooping.

[MHM Cisco World]

both network types send hello as multicast there is no different.

[Joseph W. Doherty]

The difference is DR/BDR election or not, and the time that takes to establish full adjacency.

[Joseph W. Doherty]

BTW, my earlier mention of multicast wasn't concerned with OSPF multicast but with transit multicast.  Also, that comment was NOT directed at unicast vs. multicast impact, if any, for how fast OSPF forms adjacency.

[M02@rt37 · Accepted solution]

Hello @nwekechampion 

Each OSPF interface configured as a

Point-to-Point

network type => reduce the amount of LSA's required on the network, and also within the LSDB.

[nwekechampion]

Thanks M02@rt37 .. I guess this sums it up quite nicely.. reduces the amount of lsas required, else as @MHM Cisco World mentioned, default seems to work fine regardless .. just takes a lot longer to fully establish neighborship (due to DR/BDR election)

[Joseph W. Doherty]

". . . default seems to work fine regardless . . ."

Yea, often defaults will work, as that's often one reason they are chosen to be the defaults.

Hmm, but I would caution to not "take for granted" that a default "will work fine regardless".

I suggest you understand what your options are and whether any default is suitable for best accomplishing your goals.  (Hey, and if your goal is not go beyond using defaults, except when you really, really must, that's fine too.  It's your network.)

However, for some food for thought, about defaults, consider OSPF's RFC, doesn't autocost links as Cisco does.  Will OSPF work without costing links?  Yup.  Is there benefit to costing links, maybe, but, again, it's really up to you.

Of course, Cisco's default for auto costing links is based on using a base bandwidth of 100 Mbps, and it adjusts other bandwidths based on that base bandwidth.  Can Cisco's default auto costing be left as is?  Sure, OSPF will function.  Is that best for you?  Again, that's up to you.

And so it goes with "defaults".

I suspect, when the OSPF was written, you had shared media, like Ethernet, and non-shared media used for p2p.  So, the RFC defaults assume shared media might be used between more than two routers, hence DR/BDR, to preclude the overhead of a full mesh, especially as the number of routers increase, and p2p for physical p2p links.  I.e.  defaults appears to try to be as efficient as possible for these two use cases.

BTW, on a serial p2p link, I believe you can configure network type as broadcast, and it will work!  As efficiently?  Probably not, so why do it?  If we replace the "serial" p2p link with an Ethernet link, OSPF, by default, will assume you may be dealing with multiple nodes, but since we're not, for both efficiency (and logical clarify of topology), does explicitly now defining interfaces as being p2p, make sense?  Again, that's up to the folk maintaining the network.  Possibly many would consider insuring a logical p2p physically operates as p2p is a good or even best practice, but even if so, it's your choice, to accomplish your needs and goals.

To summarize, I would recommend just considering whether "defaults" are what you want to use; i.e. don't assume they are the best choice for you just because they are the defaults.

[nwekechampion]

Many thanks @Joseph W. Doherty ,

I have just also realised that OSPF point-point network type on loopbacks has a different result to rib.. it installs full network range as opposed to broadcast which installs only /32.. possibly another use case? but any reason why it behaves that way?

[MHM Cisco World]

the LSA is reduce by only one LSA (lsa type2), 
I dont find any reason change the default. 
for the LO it have different behave because the OSPF deal with LO different than other interface. 
again in real network dont change the default.