Foreword

In 1976, when I saw my first Arpanet IMP at Digital Equipment Corporation, networks as we know them today were in their infancy. SNA, XNS, and DECnet were under early development, and packet switching versus circuit switching was the hot topic of the day. Those of us involved in the design of the switching and routing algorithms were dealing with routers (although we didn't call them that) that had 64 kilobytes of memory, data link of 56 kilobits were considered blindingly fast, and networks with 256 nodes were big enough that if you were the salesman who sold those 256 computers, you would retire fabulously wealthy.

Thirty years is a long time, and today the individual networks that make up the Internet contain thousands or tens of thousands of nodes, while the Internet as a whole contains hundreds of millions of computers. Most striking in the evolution over this human generation is that the foundations of the Internet laid down in the TCP/IP protocol suite have survived mostly intact through four or more generations of computing architectures, three complete generations of operating system technology, and an increase of five orders of magnitude in transmission speeds.

Yet, we still treat routing in packet-switched networks as a black art. Why is that?

First, designing robust, scalable distributed algorithms is hard. Despite our best intentions to make them simple, complexity creeps in to deal with the inevitable special cases, optimizations, peculiar topologies, and link technologies one encounters. Because a "fork lift upgrade" of an entire network is rarely feasible, we have multiple generations of technology present simultaneously, and we must maintain backward-compatibility with essentially no disruption to deployed services. As policies governing the routing of packets become more sophisticated, our ability to devise automated discovery and configuration procedures gets overwhelmed, and we fall back on manual configuration and performance tuning techniques. Finally, as the environment in which these networks are operated has evolved from a cooperative one where trust was implicit to one in which the network is subject to both inside and outside attack, designing and deploying routing systems that can be made secure has become an urgent priority.

Routing TCP/IP tackles this black art comprehensively. The present Volume 1 covers all the needed fundamentals of TCP/IP networks and gives you all the tools needed to understand how routing is accomplished within a single administrative region of the Internet. Straightforward ideas of packet-switched routing are presented first in the chapters on addressing and static routing. The most popular IGPsRIP, EGRP, OSPF, and ISISare covered in depth. Advanced topics in route redistribution, route filtering, and policy routing round out Volume 1.

This second edition also adds essential material on IPv6 as well as bringing all the material up to date with examples and configurations for the latest releases of Cisco IOS.

For anyone wanting a comprehensive understanding of how routing in TCP/IP networks really works, from the design principles of routing algorithms, to the evolution of addressing schemes, to the practical aspects of designing and configuring the routing of large autonomous systems, this is the book for you.

David Oran

Cisco Fellow