Summarizing (sometimes referred to as supernetting) is the process of combining multiple routes listed in a routing table and listing them as one route. The advantage to summarization is lower overhead on the router, saving memory and CPU processing time. This can be important in very large networks, which can have hundreds of routes.
Without summarization, each router must have a route to every network listed in the routing table. This might be impossible with large networks because a router may lack sufficient memory or CPU processing power.
Other possible routing solutions on a large network include creating default routes on stub networks. The default route command creates a static route that works somewhat like a default gateway, but not exactly. When a router is configured with a default route, and a packet is destined for a route not listed in the routing table, the packet is forwarded by the router to the listed default route. This will allow you to turn off dynamic routing, and the router will then forward all packets destined for a remote network to the listed default route. However, the router must be part of a stub network. What this means is that the router only has a connection to a LAN and to the default route and not a third network. If the router has a third network connection, then you cannot use default routing as your only routing for the router because a loop will occur. This will be seen with the traceroute command.
You can use a combination of static and default routing on the router to solve your routing needs. However, this doesn’t solve the problem of too many routing table entries in a router that does not have the power to handle multiple routes. If possible, you can add an access list that eliminates unwanted routes being added to a routing table. This is an option that has administrative overhead and can be difficult and cumbersome.
This Daily Drill Down will assume a network design that allows summarization points as a solution for too many routes in a router’s routing table. You will learn how to summarize contiguous blocks of networks and then configure the summary routes on a Cisco router using OSPF and EIGRP.
Designing summary routes
As mentioned, summary routes are important in a larger network. However, the problem is that routes must be in contiguous blocks to be summarized. What this means is that, for example, networks 172.16.32-63.0 are located in the same area and not spread out among different buildings or cities.
Since networks are typically installed over years, not weeks or months, finding a network design that considered a building for a possible future summarization point is usually impossible. Most networks were built on the fly and the network administrators did not even understand IP addressing; they just had a list of available networks and valid hosts. When a building needed a new network installed, the administrators just configured the next IP subnet on their list, without thinking in contiguous blocks. There is not much you can do with this type of network and unfortunately in all my years of consulting, I never once found an existing large network that was in contiguous blocks. The only time I was able to perform summarization for a large client was when they moved into a new campus environment and I was able to build the network from scratch. That was a great experience, but probably one I’ll never have again. If you ever experience being able to build a network from the ground up in a new building or campus, you may want to consider it a once-in-a-lifetime experience, so enjoy it, and design it in contiguous blocks!
Once you have your physical network in place, you can then add an IP addressing scheme to your network. If possible, tell the customer that AppleTalk and IPX are not supported on the new wiring and hardware. If you’re lucky, they’ll buy that, discard their old computers, and you can then put in a pure IP network. It’s worth a shot.
I typically would tell you that deciding on an IP scheme is done on a case-by-case project, etc. That is not really true anymore, though. The subnet mask assigned to each network is based on the needed amount of valid hosts for a particular network, but the actual IP address should be 10.0.0.0. Why? Because this is a private IP address that is not routable on the Internet. Yes, there are other private IP addresses, but the 10.0.0.0 provides the most flexibility in your network design. This private IP addressing scheme provides another layer of security for your internal network. You then would use Port Address Translation (PAT) on the boundary to the Internet. NAT/PAT is another subject, and I won’t discuss it more here. However, since you can use PAT on the boundary router, you can then use the 10.0.0.0 IP address on your internal networks and let the boundary router translate the private IP addresses to a valid IP address if a packet is destined for the Internet.
Once you have your physical network design completed, as well as an IP addressing scheme, you can configure your routers and finally, as a very last step to your configuration, configure summary routes. Do not configure summary routes until your network is up and running without any problems. Designing and implementing networks is harder than it may seem, and configuring your network with a logical step-by-step approach can save you many a headache!
A sample design
Figure A shows the physical network design that I will use to configure our summary routes.
|Obviously, this is a smaller network than you would typically have to worry about when configuring your routing table.|
Consider that the serial WAN link between the 1005A and 2500B routers is a 128K link. Terrible thought, but that makes a case for creating a summary route from the 2500B to the 1005A listing all networks connected the 2500C, D, and E routers as one route. Since the 1005A router is a small router with very low memory and CPU processing power, this now becomes a good example.
First, I need an IP addressing scheme, in contiguous blocks. Since we are using a 10.0.0.0 network, this will be very easy because of the number of subnets available with this IP scheme.
For the 2500B, C, D, and E routers, I will use a network ID of 10.1.x.0, where x represents different subnets for each physical network. The LAN off the 1005A will be in the 10.2.2.0 network, which will allow me to summarize the 10.1.0.0 network easily.
Here are my configurations for the routers:
interface serial 0: 10.2.2.5/30
interface ethernet 0: 10.2.2.33/27
interface serial 0: 10.2.2.6/30
interface serial 1: 10.1.1.5/30
interface serial 2: 10.1.1.9/30
interface serial 3: 10.1.1.13/30
interface serial 0: 10.1.1.6/30
interface ethernet 0: 10.1.1.33/27
interface serial 0: 10.1.1.10/30
interface ethernet 0: 10.1.1.65/27
interface serial 0: 10.1.1.14/30
interface ethernet 0: 10.1.1.97/27
I have used a VLSM design with 255.255.255.252 masks for the WANs and 255.255.255.224 for the LANs. This provides two valid hosts for each WAN and 30 hosts for each LAN. Figure B shows the network design with our IP addresses assigned to each interface.
|I will configure the routers and then turn on OSPF for each router.|
Here are all the routes showing in the routing table of the 1005A router.
The routes show up as RIP-found routes since I am redistributing between the 2500B and 1005A router from OSPF to RIP.
Notice the 1005A has eight subnets showing in the routing table. We can make this three routes with a summary route on the 2500B router. This will then summarize the 10.1.0.0 network as one entry to the 1005A router. All network devices with the range 10.1.0.0 through 10.1.255.254 will then have to be located off the router 2500B. You could also use the summary address 10.1.1.0 255.255.255.0, which will then summarize the network 10.1.1.0 through 10.1.1.254. This would be better if you needed a more succinct summary route. To get the above summary to work from OSPF to RIP, complicated redistribution commands were added to the 2500B router. This summary route between OSPF and RIP will not work by default.
Before we configure summary routes with EIGRP, let’s take a look at the 1005A routing table.
The 1005A can run EIGRP so no redistribution is needed between the 2500B and the 1005A. Here is how you add a summary route with EIGRP. Notice that the summary is placed on the interface for EIGRP and not under the routing protocol configuration like it is with OSPF.
The 2500B is now summarizing and sending only one route about network 10.1.1.0 to the 1005A router instead of six separate routes.
Let’s take a peek at the 1005A router, which now has in the routing table only the two directly connected routes and one summary route for network 10.1.1.0 out serial0.
The example in this Daily Drill Down showed you how to configure summary routes with EIGRP and OSPF. Typically, you would only use summary routes on larger networks, but this example is helpful in describing how to both design and implement a network design in a contiguous block.
If you have all Cisco routers, run EIGRP, and if you create your network design correctly, you can easily summarize and have a smooth-running network.