How Networks Really Work

The typical network characteristics have been misunderstood for decades, resulting in well-known Fallacies of Distributed Computing. In this introductory section, we'll revisit those fallacies and explore how they apply to modern networks.

Most network security vulnerabilities exploit the unfounded trust in the network infrastructure, from neighbors on local area networks and source IP addresses, to transport infrastructure and adjacent routers.

After briefly exploring the common networking misunderstandings, let's see what makes networks as complex as they are. We'll start with the typical networking challenges, introduce various transmission technologies, and explore the functionality we need to grow our network beyond two nodes.

Once we've identified more than a dozen challenges we have to solve to build reliable networks, we have to find a framework that allows us to build a robust system. As usual, layering and abstractions help, and we'll explore how we can use semi-independent layers to build modern network stacks.

There are two hard problems in computer science: cache invalidation and naming things. Network addressing needs to solve both problems.

This section starts with an overview of historical approaches to the network addressing, describes modern data-link- and network-layer addressing, and explores the complexities of address assignments, address scopes, and address translations.

Decades ago we were talking about bridging (forwarding on data-link layer) and routing (forwarding on network layer)... and then vendor marketers invented switches.

This section starts with forwarding terminology, continues with an overview of mechanisms one can use to get packets across the network (ranging from Token Ring Source Route Bridging to Segment Routing in IPv6), and concludes the journey with an outline of techniques used to discover paths across the network, from controller-based topology computation to transparent bridging and routing protocols.

Routing protocols are the magic that makes networks work. It doesn't matter if you're using a traditional (autonomous) network devices or an SDN controller, someone has to discover the network topology and the attached endpoints, and calculate the optimum paths across the network.

This section introduces the challenges routing protocols have to solve and describes the high-level concepts of link-state and distance-vector routing protocols. Armed with this knowledge you'll find it easier to understand the implementation details of individual protocols like OSPF, IS-IS, EIGRP or BGP.

Routing protocols became increasingly complex over the last four decades as they had to deal with environments they were never designed to support like very fast convergence, or massive amount of routing information. This section describes some of the advanced routing protocol technologies and solutions including:

• Fast failure detection and convergence
• Shared fate and trusting routing information
• Equal- and Unequal Cost Multipath forwarding

If you want to design a network that converges in tens of milliseconds you can no longer rely solely on routing protocols; the only solution is local repair. This section describes several local repair (fast reroute) technologies that can be used for temporary path repair until the routing protocol convergence process calculates the new best path:

• Prefix Independent Convergence (PIC)
• MPLS Fast Reroute (FRR)
• Loop-Free Alternate (LFA) including Remote LFA and Topology Independent LFA

Learning about the basics of computer networking is no longer expensive. This section contains links to several high-quality open-source textbooks as well as must-read books by long-time networking engineers.