OSPF Questions

[Craddockc]

Dear Community,

I am studying for the CCNP ROUTE exam hoping to pass it before the Feb deadline and had a couple of questions about OSPF operations that I was hoping you could help me out with:

1) How exactly does the router use the received Router LSA to build its OSPF tree and subsequently select the best possible route to put into the RT? I know the LSA includes all kinds of info, but at the end of the day what it really seems to boil down to is the metric right? Since OSPF calculates the cost based on incremented path bandwidth, what is the point of the router knowing about the other fields in the LSA like Link Type or even the RID of the originating router? These fields don't seem to really serve any purpose for path calculation, at least to me they dont.

2) What is the purpose of the Network LSA that the DR advertises? Why do the other routers on the multi-access network need to be explicitly told about what other routers are fully adjacent to the DR? I cant figure why thats important to the other routers.

Thanks for any help you can provide.

[Alexander Istomin · Accepted solution]

Hello.

As you know, OSPF is link state protocol. Every link state protocol based on the rule, that all routers in the same area should have same view of the network topology. Every link state protocol based on graph theory and specially on Dijkstra's algorithm. Graph is way to model pairwise relations between objects. So, somehow, each router in the area should draw network topology.

1.

For you is easy to draw the network topology, you just need understanding how your routers are located and how they are connected with each other, then you just take a piece of paper, pencil and draw it. But routers have no hands, paper and pencil, finally they can't draw. However they are "artists". LSAs is the pencil for routers, hello packets and LSU are "artist brush movement", and graph is the picture. Router ID is way to unique identify router on the picture, Link Type is way to draw lines correctly. 

2.

In OSPF there is the concept of neighborhood and adjacency. Establishing a neighborhood is a way for a router to make itself known. No neighborhood - no router. An adjacency is a state between two neighbors that allows them to exchange routes. No adjacency - no exchange of routes. You can imagine the adjacency as a pipe between two routers along which they exchange routes. If there were no DR/BDR, then on multiaccess networks, each router would have to install such a pipe with every other router according to the formula n (n-1) / 2. So, 10 routers = pipes 45, routers 20 = pipes 190, etc.

So Network LSA is a way to reduce the number of these "pipes". 

The picture below shows the situation without DR and with DR.

And yes, OSPF DR/BDR so reminds BGP router reflector...

[Craddockc]

Alex,

Thank you for the detailed and thoughtful response. Your explanation for part 1 makes sense. All the info in the LSA assists the router is building a real topological construct that contains more than just the "cost" of each link.

However, in regards to the network LSA. I guess I am still struggling with the reason why it is necessary. To me, every router on a multiaccess segment will know of every other router via the Hello packets and its "neighbor" field and at a minimum should arrive at the 2WAY state with each of them. Each router in the segment will also understand implicitly via its link type that it cant and shouldnt form any FULL adjacencies with any other routers then the DR and BDR. Between these 2 facts, I am struggling to see the value in the Network LSA. This is probably way more detail than I need for the CCNP exam but I believe in being thorough. Any thoughts?

Thanks so much again for your effort in trying to explain these concepts to me :)

[Alexander Istomin]

Again, in order for the 2 routers to exchange routes, they must be neighbors and there must be an adjacency between them. Neighborship and adjacency are different states. 2-way state is enough for 2 routers to become neighbors, but not enough for them to become adjacent. If there are no 2 different states - neighborship and adjacency, then each router would have to exchange LSA with each other router, as shown in the picture below. How to prevent this from happening on multiaccess parts of the network? The answer is the network LSA(type 2) .

DR / BDR create a fake "router", with which each other router establishes neighborship and adjacency. In this way, 2 tasks are solved:

1 - each router does not need to exchange LSAs with every other router, only with DR/BDR.

2 - and at the same time, allows the router to model(to draw) this multiaccess part of network.

If still not clear, feel free to say so :)

[Joseph W. Doherty]

BTW:

"Since OSPF calculates the cost based on incremented path bandwidth, . . ."

Cisco does, but OSPF standard doesn't tie bandwidth to cost. (Also in modern LANs, Cisco's "base" 100 Mbps = cost of 1 often needs to be manually changed.)

". . . what is the point of the router knowing about the other fields in the LSA like Link Type or even the RID of the originating router?"

Other fields have uses. For example, importance of LSA type when you hit area or AS boundaries.

RID can be helpful when "debugging" things like OSPF itself. Protocol implementations can have bugs, so it can be helpful, for example, to "know" what router is generating the information.

Lastly, cannot say for sure it's so for OSPFv2, but it's possible some items were placed into standard for some possible "future" usage. If you study the OSPF standard, and then study Cisco's implementation, especially historically, you'll find "things" Cisco has done that either aren't fully defined in standard, or are not defined, but don't conflict (at least not supposed to), with the standard.