UNIX System Administration Handbook - Evi Nemeth [210]
Link-state protocols tend to be more complicated than distance-vector protocols, but this can be explained in part by the fact that link-state protocols make it easier to implement advanced features such as type-of-service routing and multiple routes to the same destination. Neither of these features is supported on stock UNIX systems; you must use dedicated routers to benefit from them.
The common link-state protocols are shown in Table 14.2.
Table 14.2 Common link-state routing protocols
Cost metrics
In order for a routing protocol to determine which path to a network is shortest, it has to define what is meant by “shortest”.4
Is it the path involving the fewest number of hops? The path with the lowest latency? The largest minimal intermediate bandwidth? The lowest financial cost?
For routing purposes, the quality of a link is represented by a number called the cost metric. By adding together the costs of each link in a path, a path cost can be computed. In the simplest systems, every link has a cost of 1, leading to hop counts as a path metric. But any of the considerations mentioned above can be converted to a numeric cost metric.
Networking mavens have labored long and hard to make the definition of cost metrics flexible, and some modern protocols even allow different metrics to be used for different kinds of network traffic. Nevertheless, in 99% of cases, all this hard work can be safely ignored. The default metrics for most systems work just fine.
You may encounter situations in which the actual shortest path to a destination may not be a good default route for political reasons. To handle these cases, you can artificially boost the cost of the critical links to make them seem less appealing. Leave the rest of the routing configuration alone.
Interior and exterior protocols
An “autonomous system” is a group of networks under the administrative and political control of a single entity. The definition is vague; real-world autonomous systems can be as large as a worldwide corporate network or as small as a building or a single academic department. It all depends on how you want to manage routing. The general tendency is to make autonomous systems as large as possible. This convention simplifies administration and makes routing as efficient as possible.
Routing within an autonomous system is somewhat different from routing between autonomous systems. Protocols for routing among ASs (“exterior” protocols) must often handle routes for many networks, and they must deal gracefully with the fact that neighboring routers are under other people’s control. Exterior protocols do not reveal the topology inside an autonomous system, so in a sense they can be thought of as a second level of routing hierarchy that deals with collections of nets rather than individual hosts or cables.
In practice, small to medium sites rarely need to run an exterior protocol unless they are connected to more than one ISP. With multiple ISPs, the easy division of networks into local and Internet domains collapses, and routers must decide which route to the Internet is best for any particular address. (However, that is not to say that every router must know this information. Most hosts can stay stupid and route their default packets through an internal gateway that is better informed.)
While exterior protocols are not so different from their interior counterparts, this chapter concentrates on the interior protocols and the daemons that support them. If your site must use an external protocol as well, see the recommended reading list on page 371 for some suggested references.
14.3 PROTOCOLS ON PARADE
Several interior routing protocols are in common use. In this section, we introduce the major players and summarize their main advantages and weaknesses.
RIP: Routing Information Protocol
RIP, defined in RFC1058, is an old Xerox protocol that has been adapted for IP networks. It is the protocol used by routed. RIP is a simple distance-vector protocol that uses hop counts as