Fundamentals of Default Routes

When a router is connected to the Internet, a default route is immensely useful. With a default route, the router needs to only recognize destinations that are internal to its own administrative system. The default route will forward packets destined for any other address to the Internet service provider. This negates the necessity of running Border Gateway Protocol (BGP) with the service provider to learn all of the prefixes in the Internet route tablea table which consists of well over 100,000 prefixes, and might soon be approaching 200,000. In dealing with large route tables, topology changes are an even bigger concern than the demands on memory. In a large network, topology changes will occur more frequently, resulting in increased system activity to advertise and process those changes. Using a default route effectively "hides" the changes of more-specific routes, making the network to which the default points appear more stable from the point of view of the router using the default route.

Default routes are also useful on a smaller scale, within single autonomous systems. The same benefits of decreased memory and processor utilization can be gained in smaller networks, although the benefits decrease as the number of routes decreases.

Default routes are also very useful in hub-and-spoke topologies, such as the one in Figure 12-1. Here, the hub router has a static route to every remote subnet. Entering new static routes in the hub router when a new subnet is brought online is a fairly trivial administrative task, but adding the routes to every spoke router might be much more time-consuming. By using default routes at the spoke routers, only the hub needs entries for every subnet. When a spoke router receives a packet for an unknown destination, it will forward the packet to the hub, which can, in turn, forward the packet to the correct destination.

Figure 12-1. Default routes greatly simplify the administration of static routing in a hub-and-spoke network.

The spoke routers in Figure 12-1 are more correctly called "stub" routers. A stub router has only a single connection to another router. The routing decisions become very simple in such a device: The destination is either one of the router's directly connected networks (stub networks), or it is reachable via its single neighbor. And if the single neighbor is the only next-hop routing choice, the stub router has little need for a detailed route table. A default route is usually sufficient.

As with other summary routes, the trade-off with default routes is a loss of routing detail. The stub routers in Figure 12-1, for instance, have no way of knowing whether a destination is unreachable. All packets to unknown destinations are forwarded to the hub router, and only then is reachability determined. Packets to nonexistent addresses should be infrequent in a network. If for some reason they are not, a better design choice might be to allow the stub routers to run a routing protocol and learn routes from the hub so that unknown destinations can be determined as soon as possible. The design choice for you to make in a network such as the one in Figure 12-1 is whether it is more economical to forward packets with unknown destinations to the hub router, which can then drop them, or whether it is more economical to run a dynamic routing protocol between the hub and stub routers just to drop packets to unknown destinations at the stub routers. Although the resource and operational costs of running a dynamic routing protocol are usually small, the default route is still more likely to be the best choice.

Another problem with loss of routing detail is shown in Figure 12-2. These routers form a nationwide backbone, and large local networks are connected to each of the backbone routers. The Los Angeles backbone router has default routes pointing to both San Francisco and San Diego. If Los Angeles must forward a packet to Seattle and has only the two default routes, it has no way of knowing that the best route is via San Francisco. Los Angeles might forward the packet to San Diego, in which case the packet will use a small portion of some very expensive bandwidth, and will incur some unnecessary propagation delay, before it belatedly reaches its destination. Using default routes on this backbone is a bad design decision,^[2] but it illustrates how hiding route details with a default route can lead to suboptimal routing.

^[2] Having each backbone router advertise only a default route into its local network, on the other hand, can be a very good design choice, limiting the size of the local route tables.

Figure 12-2. If the Los Angeles router knows only default routes pointing to San Francisco and San Diego and has no more specific details about the topology behind those two routers, it cannot route efficiently.