Illustrated TCP/IP by Matthew G. Naugle Wiley Computer Publishing, John Wiley & Sons, Inc. ISBN: 0471196568   Pub Date: 11/01/98

Chapter 92
IP Routing

We had to go through the IP addressing section in order to understand routing. Hopefully, this section will be a lot more comprehensible. Packets are routed based on the address that is in the packet. Routers read this information and determine the best path known as the next hop. A packet switched network (compared to a circuit switched network) is based on a unit of information (known as a datagram) and its ability to make its way through the network to the destination. The datagram may be routed locally (the destination is on the same subnet as the originator) which is known as direct routing, or it may invoke the use of a forwarding device such as a router if the destination is remote (on a different subnet than the originator). The latter is known as indirect routing, which infers hierarchical routing. A datagram that is sent may invoke both direct and indirect routing.

Why not just have one large flat network? Place everyone on the same network. ATM tried to do this as well as switches and bridges. Eliminate indirect routing completely. Flat networks do have their place: in small networks or WAN protocols or to extend a subnet through switches or bridges. With the current suite of network protocols, a large flat network is inefficient (it does not scale well), especially when you estimate the millions of addressable stations that are attached to it. And the protocols that currently run on networks are broadcast oriented. This means the network allows for multiple stations to be attached and grouped to a single network, and these stations see all data on their network no matter who sent it and who it is for. The protocols were built for shared environments. These networks were invented before the advent of switches and routers. Also, when stations need to communicate, the initial communication could be sent in broadcast mode. Communication between certain devices (routers) is always done in broadcast or multicast. This is a special type of packet that enables all stations to receive the packet and hand it to their upper-layer software to filter or process. As you scale for growth, a network cannot remain flat. There must be some sort of hierarchy to allow for efficiency.

Not all stations need to see each other. As a network scales, it must maintain its manageability. To make any network more manageable, it will be split into many networks called subnets (virtually any network today, whether split or not, is called a subnet). To make these subnet networks manageable they will in turn be split further into sub-subnets. The interconnection of these subnets is accomplished by forwarding devices known as routers. Routers enable data to be forwarded to other networks in a very efficient manner. It will always be easier to manage many smaller networks than it will be to manage one large network. Also, broadcast data stays on its network or subnet. It is not forwarded by routers (exceptions occur and they will be noted in those sections, such as DHCP or all subnets broadcast).

In order for routers to forward data to other networks, they use special protocols (known as routing protocols) to enable them to internally draw a map of the entire internet for the purposes of routing. To accomplish this, there are two types of protocols used: Interior Gate-way Protocols (IGPs) and Exterior Gateway Protocols (EGPs). The Exterior Gateway Proto-col used with IP is known as Border Gateway Protocol (BGP). The IGPs that I will explain are known as the Routing Information Protocol (RIP and RIP2) and Open Shortest Path First (OSPF).