A name indicates what we seek. An address indicates where it is. A route indicates how we get there.
Unicast routing is a fundamental enabling technology for the cloud-native data center. And though unicast routing is as old as the Internet, its application in the data center has led to some head scratching. Some of the head scratching led to enhancements and change to routing protocols in adapting them to data centers. But some of the re-examinations of old practices resulted in erroneous pronouncements. These errors are either based on an excessively strict application of theory to the Clos topology (saying, for instance, that link state protocols are too chatty) or on popular misconceptions (for instance, that BGP is too slow).
Another barrier to proper routing within the data center is that, for many network operators raised on building bridged networks, routing seems an arcane art. So this chapter is an attempt to clear the cobwebs in the path to a better understanding routing and routing protocols, specifically with their application to Clos topology. With the rise of technologies such as Kubernetes, even non-network engineers have to take a look at routing; Kube-router and Calico require an understanding of routing and routing protocols.
This chapter should help you answer questions such as:
How does routing work?
What are the types of routing protocols?
What is the fit of routing protocols to Clos topologies?