Cisco Support Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Announcements

Welcome to Cisco Support Community. We would love to have your feedback.

For an introduction to the new site, click here. And see here for current known issues.

New Member

OSPF Design problem/question

I have the following situation:

G

||

||

A====B====L

{{{{}}}}

{{{}}}

{{}}

{}

||===H --------C (T-1)

||

D------------------- INTERNET (T-3)

||

||

K------E (T-1)

|

|

F (T-1)

=, || (LAN, 100 Mb connections)

{} (LAN, 1Gb connections)

-- (WAN connections)

Current configuration:

A, B, C, D, L are 6500s.

E, F, G, H, K are 2600.

A,B,D, L are Gigabit LAN connection within campus network.

G, H , K are 100 Mb connections within campus network

C,E,F are T-1 connections to the remote locations.

D is connected to T-3, the main exit to the internet . Stub network

Everything is one big OSPF network, under area 0.

The network is class B, and it is not contigues. It is being advertised by our ISP as one big class B.

Total of 248 routes. Routes are advertised as follows:

A: 38

B: 12

C: 12

D: 22+4

L: 14

G: rest of the routes ( sink to Interface Null)

Our connections between remote location C is going to change. Therefore, I would like to redesign the OSPF network.

Our ISP will bind us to its LSS circuit through D connection for the entire campus, except C network. They will bind the network C to its LSS circuit directly.

So now I will have

(only middle portion of the diagram)

INTERNET--(LSS-14 routes)--C, totaly separted from, D--(LSS-235 + routes)--INTERNET(Entire campus, except C)

They will split our network into 2 parts and advirtise 14+ routes separetly from 235 + routes, as class C networks via BGP. So now my routers C and D will also be running BGP and redistributing the routes.

What is the best way to efficiently design this network?

I was thinking to split the network into 2 different OSPF ASs. C being one and having only area 0. D being second AS with A, B, D, L, G being area 0, and K, E, F being area 1 (totally stub network). I wanted to split A, B, D user sides (Vlan interfaces) in separate areas, but due to unability to summarize them (overlapping of route summaries because networks are not contigues), I would have to stick to one big area 0 on the main campus.

Since I am not experienced with OSPF design, I would request possible ideas, comments, recommendations, etc.

Thanks,

sp

1 ACCEPTED SOLUTION

Accepted Solutions
Gold

Re: OSPF Design problem/question

PE/CE refers to an MPLS VPN. If you're service provider hasn't told you you're running an MPLS/VPN through their network, then you probably aren't (or they don't give you the option of running OSPF through their network using BGP/OSPF PE/CE capabilities).

If they are advertising your routes for you, then you don't need to worry about how to advertise them. A static to null0 is just what it sounds like:

ip route x.x.x.x x.x.x.x null0

which creates a route that will never be removed from the routing table, hence it's "nailed up." Advertising routes created in this way is a common practice among service providers. It sounds like you are only receiving routes through BGP, though, and not sending them, so this isn't an issue.

Finally, Is it 238 routes, or 238 routers? A well designed network can handle 200 to 300 routers in area 0, with no problems. "Well designed" is where the problem lies, though, and it's hard to actually define that term in this short of a space. A good place to start would be the book _Advanced IP Network Design_, on Cisco Press.

If it's 238 routes, you're fine. The other area, with 14 routes (or routers?) is fine in either case.

:-)

Russ.W

9 REPLIES
New Member

Re: OSPF Design problem/question

Let me rephrase it:

I would really appreciate if someone would be kind enough to present me with some ideas, or make comments refering to my proposed solution.

sp

Bronze

Re: OSPF Design problem/question

It's hard to get a good idea of the current and proposed designs with ASCII diagrams. Can you quickly whip something up in Visio and attach it to this thread?

New Member

Re: OSPF Design problem/question

Next post

New Member

Re: OSPF Design problem/question

Here are the attached diagrams.

Net1 is current layout.

Net2 is future layout after LSS circuits are in place.

Thanks for the help.

sp

Gold

Re: OSPF Design problem/question

Since you're running BGP through the service provider, you don't really have much option other than splitting the two sites up into two different processes or "autonomous systems," unless you're running PE/CE with the service provider, which you don't mention (?). The sites themselves aren't very big (or at least it doesn't look that way from the drawings), so I'd just put each on in area 0.

Creating a stub area at the one site certainly isn't going to hurt anything, though, and it will help you scale the network in the future, as more possible sites and areas are added. You don't need to summarize towards the core to create a stub area at the other site, either--you can just create the stub area, and let it advertise all the reachable destinations into the core, which is fine. The routers in the stub area will have less routing information this way.

My main questions would be around the interaction between BGP and OSPF at the edge of the service provider network. The primary consideration is how you're going to get your routes into BGP at that edge--by summarizing and advertising the ospf route, or by nailing up a static to null0, and advertising that into BGP? I'd probably go with nailing up the static to null0, and advertising that.

If you get some other back door link in the future, things will be more complicated. Considerations there would be the ospf wait for bgp convergence feature, to make OSPF and BGP interact a little more nicely at that service provider edge, and how to handle the back door link in the area structure--put it in area 0 is the more likely answer, but there are a couple of things there to consider.

For right now, I think you're on the right track with where you're going, though.

:-)

Russ.W

New Member

Re: OSPF Design problem/question

As I said earlier, I am not that experienced with OSPF, so please forgive my ignorance.

I have no idea what you mean by PE/CE (please provide some links that explain the terms/situation). However, my understanding from our last meeting with UEN (my ISP), is that we would, probably, have two different ASs.

Regarding BGP-OSPF interaction, my understanding is that instead advertising one class B network, they will advertise us on individual class C networks basis, and implement summarization where possible. However, you are suggesting nailing up the static to null0. Again, sorry for my ignorance, I have no idea what you meant with this. Would you, please, explain it to me (or post some links to white papers), and benefits of doing it that way.

I don't know how you, experienced engineers, determine the size of the network, as to regard of creating multiple OSPF areas. And I don't know what you consider a small vs. large network. On my main campus,there are around 238 routes, and on the remote campus (router C) there are only about 14 routes. I am not sure if 238 is too big for one area 0, or should it be split in several areas.

Thanks for the help,

sp

Gold

Re: OSPF Design problem/question

PE/CE refers to an MPLS VPN. If you're service provider hasn't told you you're running an MPLS/VPN through their network, then you probably aren't (or they don't give you the option of running OSPF through their network using BGP/OSPF PE/CE capabilities).

If they are advertising your routes for you, then you don't need to worry about how to advertise them. A static to null0 is just what it sounds like:

ip route x.x.x.x x.x.x.x null0

which creates a route that will never be removed from the routing table, hence it's "nailed up." Advertising routes created in this way is a common practice among service providers. It sounds like you are only receiving routes through BGP, though, and not sending them, so this isn't an issue.

Finally, Is it 238 routes, or 238 routers? A well designed network can handle 200 to 300 routers in area 0, with no problems. "Well designed" is where the problem lies, though, and it's hard to actually define that term in this short of a space. A good place to start would be the book _Advanced IP Network Design_, on Cisco Press.

If it's 238 routes, you're fine. The other area, with 14 routes (or routers?) is fine in either case.

:-)

Russ.W

New Member

Re: OSPF Design problem/question

Thanks for the advice. I appreciate your help. I meant "routes" not routers, so one area would be sufficient for the campus network.

Not urgent, but when you do find some time, would you please explain me what happens when you issue a static route with null0, as you did? What is the destination?

On my sink router, all my unused class C networks, have static route to Null0. However, I was under impression that in this case packets do get dropped (insted roaming through the network). If that is the case, how would they (my ISP) advertise my networks with null0. Doesn't that mean that packets are dropped?

Once again, thanks for the prior explanations.

sp

Bronze

Re: OSPF Design problem/question

SP,

Using static routes and pointing them to Null0 gives you control over whether or not your router will route or drop a packet based on the route configured.

For example lets say you have a router that specifically knows about the 10.10.1-100.0 255.255.255.0 and you would like to only advertise one route in your dynamic routing protocol process to a neighbor.

You could configure a static route "ip route 10.10.0.0 255.255.0.0 null0" and then redistribute this route into your dynamic protocol.

What happens is when your router receives a packet destined to 10.10.x.x it will look in its routing table for the most specific route it has. So lets say its destined to 10.10.50.5/24. Your router already has a specific route to that subnet so it will route the packet. But lets say you got a packet destined to 10.10.150.x/24 your router will not find a specific route for the 10.10.150.x/24 which would cause the router to match the static route pointed to null0 zero which means drop.

I hope that gives you a little bit of an understanding of routing to null0. The idea behind null0 is that you will have more specific routes in your table and the null0 route is a more of a summarized route. That way your router will route packets to specific networks it knows about and drop everything else that would match the null0 route.

Daniel

186
Views
0
Helpful
9
Replies