Designing Implementing an OSPF Network | OSPF Network Design Solutions

Table of Contents

CAMPUS to WAN Routing

Now that all WAN routes are available on the campus IGRP backbone, it will be necessary to advertise routing information to the WAN routers such that all campus subnets can be reached. This can be accomplished in two ways:

• Redistributing IGRP routes into OSPF

• Generating a default route into OSPF

All known IGRP subnets can be redistributed into OSPF at the Hub router with the following:

    TENN#    router ospf 5775    redistribute igrp 10 metric 100 metric-type 1 subnets

metric 100 is an arbitrary default metric that will be attached to IGRP routes redistributed into OSPF.

metric-type 1 will make redistributed IGRP routes external Type 1, which will allow the OSPF spoke routers to add individual link costs in order to calculate OSPF metrics.

subnets is necessary to allow subnets of natural class B address 172.17.0.0 to be redistributed into OSPF.

You should take note that the IGRP “gateway of last resort,” default route or 0.0.0.0, will be automatically redistributed into OSPF because it appears as an IGRP route on the Hub router. This default route will be propagated to all frame-relay routers as an external 0.0.0.0/0 route.

When generating a default route into OSPF, because all spoke routers only have a single path (one PVC) out to the WAN, all destinations, which are not locally connected, would have to traverse that path. A default route (0.0.0.0/0.0.0.0) can be sent from the ASBR Hub router in lieu of specific subnet routes. This is the preferred method in this case, since the routing tables on remote routers become smaller, and potential routing loops which sometimes result from two-way redistribution can be avoided altogether. The syntax that follows demonstrates this procedure.

    TENN#    router ospf 5775    default information originate metric 100 metric-type 1

Notes:
In order to use the default route method, it is necessary to enable classless interdomain routing (CIDR) on all OSPF routers. This is done using the “ip classless” command on all routers:
Router# ip classless

Without this command, remote routers will not use the default route as a possible path for any destination networks, which are subnetted out of the “native class” 172.17.0.0 network.

Refer to Figure 7-29 to see the implementation of the techniques covered in this section for Campus-to-WAN routing.

Case Study Conclusions

Although the Terrapin OSPF network design was fairly simple in terms of IP addressing and OSPF architecture, integration into the IGRP network presented a number of challenges. Adding OSPF into existing networks running other routing protocols is often a difficult task, and must be carefully planned out; otherwise sub-optimal routing or even loops may occur.

Frequently Asked Questions

Q—What is the benefit of a one-layer distributed network design?

A—They are good for smaller networks and are useful from a survivability aspect if you plan to distribute servers throughout the network. The downside is a tendency to have duplicated functions at the various sites, which results in higher costs.

Q—I have a server farm type of LAN within my networking environment. How should I design the network to maximize their placement?

A—The standard hub and spoke network design would be good for locations having a server farm. You might also want to consider using the higher bandwidth LANs (fddi and fast Ethernet) to connect the servers.

Q—Why shouldn’t I put hosts and users on the backbone of my network?

A—To facilitate effective routing at the backbone, users should not be directly connected to it. Ideally, about 80 percent of LAN traffic should remain there. By following the guidelines, you will increase the backbone’s reliability, facilitate proper traffic management, and be able to easily plan for backbone equipment or bandwidth upgrades.

Q—Where can you get OSPF software?

A— The first thing that comes to mind is Cisco routers. However, there are other places to find the software such as other router manufactures. In addition, Merit Networks of Ann Arbor, MI currently maintains a program known as GATED. This program is among other things a routing daemon for Unix platforms and it contains OSPF. The original implementation of OSPF has been incorporated into GATED. For additional information on these two sources, refer to the following sites: http://www.cisco.com and http://www.gated.com.

Q—How can I learn more about networking and OSPF in particular?

A—I would recommend two mailing lists. The first is a mailing list regarding Cisco networking equipment as a whole. You can join it by sending a subscription request to: cisco-request@spot.colorado.edu.
The other mailing list that I would recommend concerns the OSPF protocol as a whole. The OSPF Working Group has a mailing list that holds discussions on various OSPF topic. You can join it by sending a subscription request to: ospf-request@gated.cornell.edu.
You can also go to the IETF and learn more about the IETF Working Groups and the areas of networking that they monitor and such: http://www.ietf.com.
Cisco also has a very extensive Web page that contains an impressive amount of information on networking, although to access the majority of it, you need a Cisco Connection Online (CCO) account. This information can be found at http://www.cisco.com.

Q—I need an explanation of OSPF link types. Could you please summarize and explain the differences between the following:

• OSPF router links

• OSPF net links

• OSPF summary net links

• OSPF exterior links

A—RFC 1583 describes what those links are. The following information is in RFC 1583:

• List of router link advertisements. A router link’s advertisement is generated by each router in the area. It describes the state of the router’s interfaces to the area.

• List of network link advertisements. One network links advertisement is generated for each transit multi-access network in the area. A network link’s advertisement describes the set of routers currently connected to the network.

• List of summary link advertisements. Summary link advertisements originate from the area’s area border routers. They describe routes to destinations internal to the Autonomous System, yet external to the area.

• List of external routes. These are routes to destinations external to the Autonomous System that have been gained either through direct experience with another routing protocol (such as EGP), through configuration information, or through a combination of the two (for example, dynamic external information to be advertised by OSPF with configured metric). Any router having these external routes is called an AS boundary router. These routes are advertised by the router into the OSPF routing domain via AS external link advertisements.

• List of AS external link advertisements. Part of the topological database, these have originated from the AS boundary routers, and they comprise routes to destinations external to the Autonomous System. Note that if the router is itself an AS boundary router, some of these AS external link advertisements have been self-originated.

Q—I want to run OSPF over ISDN (DDR). How can I suppress that the connection is established for every hello packet? Does snapshot routing work with OSPF, or just with distance vector protocols?

A—Snapshot will not work with OSPF because it is a link-state protocol. However, OSPF over DDR links is supported in Cisco IOS 11.2. This feature enables you to suppress hellos and updates after the updates and hellos are passed initially.

Q—What is the best way to implement an IP default network (0.0.0.0) in a mixed RIP/OSPF network? I have inherited a network that has an OSPF backbone and is redistributed into RIP on the boundary routers. There are static routes to 0.0.0.0 scattered throughout the network, and they are redistributed in OSPF and RIP. In fact, there is a static route to 0.0.0.0 pointing to the next hop for Frame Relay defined on every router. There are IP default network statements on every local campus router, and there is an OSPF default-information originate statement. What is the best, most fault tolerant methodology to introduce a simple clean default route in this environment?

A—Remove all the static routes. Watch where you dynamically get a default route. Make sure you have the default-info originate command on the correct router that you want to generate the 0.0.0.0 route. Let this go to every area dynamically. When you redistribute OSPF into RIP, this route will also go dynamically. Remember that you cannot originate default on every OSPF router: 1) Stub areas do this automatically; 2) Only ABR can generate this by using the default-information command.

Q—Can OSPF give me full connectivity for IP in a partially-meshed Frame Relay network? Or, will I have to configure subinterfaces?

A—OSPF has a feature called point-to-multipoint interfaces to easily allow full connectivity over a partially meshed Frame Relay network in Cisco IOS 11.0 and later. It could be done before Cisco IOS 11.0, but it required the hub router to be the Designated Router and some map statements on each spoke router to all the rest of the spoke routers through the hub router.

Q—Can the ip ospf network point-to-multipoint command be used with Frame Relay subinterfaces?

A—Yes, that is exactly what it was designed for.

Q—Can IPX be routed using OSPF?

A—No, OSPF is for IP. NetWare Link Services Protocol (NLSP) is Novell’s answer to link-state routing protocol for IPX.

Q—I have a serial link between Router A and Router B. They will use an ISDN link for bandwidth-on-demand and backup, as indicated here:

    backup interface bri 0    backup load 25 5    backup delay 10 60

The serial link has OSPF and IPX configured. Is it possible to have all the IPX and OSPF perimeters transferred to the ISDN link when the serial reaches 25 percent, or must I configure OSPF and IPX on the BRI?

A—You need to configure both the IP and IPX parameters on the BRI interface also. After all, bringing up another link means another physical path to the destination and this path must contain the IP and IPX information before the router can put IP and IPX traffic onto this link.

Q—Does OSPF support secondary addressing? Is there anything special I have to configure?

A—Yes, secondary addressing is supported. The secondary address needs to be in the same area as the primary interface. In addition, OSPF cannot be configured on a secondary interface without being enabled on the primary.

Q—Can I redistribute interior BGP routes into OSPF?

A—Yes, but it is really not allowed and is strongly discouraged. Otherwise, you might cause routing loops.

Q—My ASBR router is running OSPF as well as BGP. The router knows about my network’s IP addresses through OSPF—the sh ip route command shows all my OSPF routes. But the BGP process does not exchange these routes with its peer, even if I use the BGP command: network <number> <mask>, where <number> <mask> are the aggregated IP addresses and corresponding masks of networks known by the ASBR router’s OSPF process. If I redistribute the OSPF routes into the BGP process using the redistribute ospf <id> route-map <ospf-to-bgp> command, then only BGP will start redistributing my networks to its peer. I have taken care to use the proper access list associated with the redistribute command. The Cisco manuals recommend not to redistribute IGP into BGP, however, and state that the better way to do this is to use the BGP network command, which does not seem to work for me. How can I resolve this problem?

A—For the network command to work, you need to have the exact route specified in it contained in your routing table. Make sure that your IP routing table has the exact routes that are mentioned in the network statement. Refer to BGP Technical Tips at: http://www.cisco.com/warp/customer/459/ 18.html or the Cisco troubleshooting engine which is found at: http://www.cisco.com/diag/te_start.html.

Q—Can you use the distribute-list out command to filter static routes that are being redistributed into OSPF? I have a network running OSPF. On some of my routers, I have static routes that are being redistributed into OSPF; however, I do not want all of the static routes to be redistributed. I used the distribute-list out command, and this appears to have worked, but I have found that if I add another access-list command to permit an additional static route to be redistributed, the new access-list has no effect until I remove the distribute-list out command from the OSPF routing process and then re-insert it.

A—What you have done is fine. You can also use the clear ip ospf redistribution command to refresh the redistribution process.

Q—When implementing an OSPF network, what are the advantages and disadvantages in establishing Area 0 for the whole network?

A—Generally, it depends on the total number of routers in the network and the topology of the network. If you are going to have fewer than 40 routers in the network, you should be able to get away with having all routers in area 0. For larger networks, you will want to subdivide your network to break it up into areas.

Q—Can you set an OSPF dead interval timer in Cisco IOS 11.0 on a 4500 series router? I configured a core router (FDDI-attached) with the same configuration as a local router, but the users could not see out of their LAN. Both of the interfaces configured are Ethernet. There was an OSPF hello interval statement and network statements in the one router, and both had area statements.

A—Yes, you can change the dead interval using the ip ospf dead-interval command. If this timer is not manually set, it will take a value of four times the Hello interval, by default.

Q—What is the command to enable the serial interface learn routes only (listen) and not send the updates? I am running OSPF in a Cisco 7000 router.

A—You will not be able to do this. The command passive-interface serial x with OSPF will disable both incoming and outgoing routing updates. The command distribute-list out cannot be used, unless this is an autonomous system border router (ASBR), and you only want to filter external routes (from other routing protocols).
On the other router, the most you can do is filter incoming updates from the serial link with distribute-list in. This will affect only those routers coming into that particular router’s routing table. It will not alter the OSPF database.
Thus, that router will pass the LSAs on to its other neighbors, so even this is not a very good solution. You would need a distribute-list in on each subsequent router to block the LSAs from getting into each router’s routing table.

Q—If area1=NJ, area2=Delaware, and area0=NYC, will routing ever take place between area 1 and area 2 without traversing the backbone (as it will be the shortest path)?

A—In the migration of RIP to OSPF, there seems to be a case where two non-backbone areas are going to be connected, such as area 1 to area 2. This is in place for redundancy. In OSPF, all areas must touch the backbone area 0. You can, however, use virtual links to get from area 1 to the backbone and then to area 2. The virtual links are tunnels to the backbone.

Q—What is the recommended maximum number of routers in an OSPF area, specifically the backbone area?

A—It depends on how stable your network is. If it is extremely stable, with no flapping links, you can get by with more routers in an area. 40 is a conservative estimate.

Q—What is the correct wildcard mask for an OSPF network with mask 255.255.255.252? We are using Class C network addresses, on Loopback interfaces for DLSw, subnetted 255.255.255.252. From the three proposals that follow, what is the correct wildcard mask to use on the OSPF network definition statement router ospf 1234:

1. network 192.168.1.4 0.0.0.2 area 0 (two hosts)

2. network 192.168.1.4 0.0.0.3 area 0 (two hosts and broadcast)

3. network 192.168.1.4 0.0.0.4 area 0 (network, two hosts, and broadcast)

A—The third wildcard mask is the best method. For the loopback interface, you can use a 255.255.255.255 mask to save address space because OSPF supports host-route. To answer your question, if your loopback is 192.168.1.9 255.255.255.252, then under OSPF you would have: network 192.168.1.8 0.0.0.3 area 0. In other words, #2 in your question is the correct choice.

Q—How does OSPF handle multiple exit points? I am running OSPF between areas and RIP within areas. I also have multiple exit points to some destinations. How do I tell OSPF to use the preferred path? Do I need static route statements to force it?

A—OSPF will use the lowest cost to the end destination. You can find the total cost by adding up all the costs on individual links to the exit points. If you want, you can manipulate the OSPF costs with the ip ospf cost command. The administrative distance of RIP is 120 and OSPF is 110.

Q—Is there anything I should look out for with Dial Backup using OSPF?

A—It will work well as long as you obey the general rules of OSPF, such as never dial between areas (from area 1 to area 2, for example). Also, make your backup delay intervals long enough so you don’t end up causing route flaps in the OSPF, which will cause routing storms.

Q—If I inject some external routes into OSPF, can I then summarize those routes into one route on some router downstream in the network? It seems that the only way to summarize the external routes is by summarizing on the router that first injected (redistributed) the external routes. I am using the summary-address command. I do not think the area range command will work either because this summarizes only internal OSPF routes.

A— You must summarize the external routes at the ASBR router (the router where the redistribution of external routes into OSPF is taking place). Please see the OSPF Design Guide for details.

Q—While testing OSPF, I ran OSPF on one of my routers and used the OSPF default-information originate metric-type 1 command. This same router had a static route ip route 0.0.0.0 0.0.0.0 xxx.xxx.xxx.xxx, where xxx.xxx.xxx.xxx was the opposite directly connected interface on s0 of my test router. When I shut down s0, the default route generated by the default-information originate command disappeared. Does that mean that the OSPF default-information originate command polls the interface defined in the static route?

A—If the next hop in a static route is unreachable or down, the static route will no longer be installed. In addition, because that static route is no longer in the router, OSPF is not originating default. For the router to originate a default without any regard to the availability of a default route in the router, use the command default-information originate always.

Q—When using OSPF, is area 0.0.0.0 the same as area 0? Should the backbone always be area 0?

A—Yes. Area 0 is also displayed as area 0.0.0.0. Yes. Area 0 is the backbone area, and it is mandatory.

Q—Do Cisco routers support OSPF for secondary addresses? We would like to run OSPF for both the primary and secondary addresses, but have that interface as passive. A second scenario (but not yet needed) is OSPF on the LAN with OSPF neighbor as primary IP address, with neighbor as secondary IP address—is that possible?

A—Cisco routers treat secondary addresses as stub networks under OSPF. They also do not form adjacencies on secondary addresses. It has to be the primary for that to happen.

Q—Is it possible to create an OSPF virtual link through two or more areas to connect to area 0?

A—That is not a supported function according to the RFC. You cannot create an OSPF virtual link across more than one area.

Q—How do I set the hub router on a hub and spoke Frame Relay network to be the designated router (DR) in an OSPF routing environment?

A—Most hub-and-spoke architectures do not have any broadcast capability to emulate shared media between remote nodes, but instead are made up of a set of point-to-point links converging on one central router. In any point-to-point topology, even with a large set of neighbors, the DR concept is irrelevant and not used. The routers will establish a complete adjacency without electing a DR. In many cases, the above might require that each subinterface be set up as an explicitly point-to-point subinterface and/or the use on a per-interface basis of: ip ospf network point-to-point. If, for some reason, you want to establish DR/BDR relationships on point-to-point links, you may still implement ip ospf priority < number >, where < number > is greater than one and applied on the local interfaces to the hub router. In such a topology, however, this is not recommended.

Q—How do I advertise a single summary route to other OSPF areas? I have subnetted a Class C network no. With a 255.255.255.252 mask for my Frame Relay backbone. For each of my point-to-point connections, I am using a subnet, and on the Frame Relay backbone all subnets are contiguous. How would I summarize all routes using OSPF to advertise a single summary route to other OSPF areas? For example, based on a 207.105.207.0 network no., is the following correct?

    Network no.: 207.105.207.0    Subnet mask: 255.255.255.252

A—Summarize all addresses between 207.105.207.4 through 207.105.207.44 with the following: area 12 range 207.105.207.4 255.255.255.47. The area range command summarizes a block of addresses in an area to the backbone. This command should be configured on the backbone area border router. For example, if you were to summarize a block of addresses 131.108.32.0-131.108.47.255 in area 12 to the backbone; the required command would be area 12 range 131.108.32.0 255.255.240.0.

Q—If a router (router 1) has a default route specified to be to a router (router 2, which is not talking OSPF) on a network on which router 1 has an interface, and router 1 has default-information originate set so that it propagates a default route, will the default route advertisements have a forwarding address set to the address of router 2 so as to avoid ICMP redirects?

A—The OSPF updates should contain database information including the IP address of the location of the default router. The OSPF router receiving the update should independently decide the best “next-hop” to get to the network where that address resides. If it is directly connected to the subnet in common with router 1 and router 2, it should send the packet directly to router 2 even though router 2 does not speak OSPF. Unlike classic distance-vector protocols, the “next-hop” address is independent of the advertising router.

Q—When using floating static routes on a link backup over ISDN, how can I ensure that the floating route stays active until the OSPF table is built?

A—Floating static routes have administrative distances greater than the other routing protocols (higher than 120). This means that a route learned via any routing protocol will take precedence, regardless of the metric. When the link is restored, the router will use the OSPF route as soon as it is available. The backup link will stay up until the backup delay timer expires, but IP traffic will use the OSPF route.

Q—I have a network (Frame Relay) in which every remote site has a primary and a secondary link. I want the secondary link to be used only if the primary becomes inaccessible; OSPF is my routing protocol. When I do trace route, the path taken to reach the destination is through the secondary link. Why does this happen?

A—I think you mean that both links are connected to the same router. The best way is to use different ip ospf cost on the links or use static routing for the backup link with a higher distance than OSPF.

Q—How do I resolve discontiguous networks?

A— You can use secondary IP addresses to link the address space or you can use OSPF EIGRP, IS-IS or BGP4 with auto summarization turned off.

Q—In 9.1, why must the neighbor command be used when running OSPF over X.25 networks?

A—You need the neighbor command to make OSPF work on X.25 in 9.1. In 9.21 and later, at OSPF level, an X.25 network can be configured to be a broadcast network, and OSPF would treat X.25 as a broadcast network only. X.25 maps with the broadcast keyword would be needed to make it work.

Q—How can an OSPF default be originated into the system based on the existence of certain external information (i.e., routes learned from some exterior protocol) on a router which does not itself have a default?

A—OSPF will generate a default only if it is configured using the command default-information originate and if there is a default network in the box from a different process.

Table of Contents