1.1 Network Administration

In the past two decades the number of networked computers in the world has grown at an astonishing rate. In the mid 1980s when the personal computer became the hot new item, most people had no concept of connecting these machines. Today, a mere 20 years later, it is hard to imagine an organization with more than a few computers that does not connect them in some fashion. More often than not, these networks are also connected to the global Internet, allowing connectivity to any other Internet-connected machine anywhere else in the world.

When networking technology was not so widely used, there was little need for network management. There simply wasn't much to manage. And when things did go wrong, it was usually the hardware at fault. Today the hardware is much more reliable, and the problems are often caused by bad software or malicious users. At the same time, the need for reliability has increased. More people now rely on the network in order to accomplish their jobs; some networks even support life safety equipment. These two influences, the greater need for reliability and the fact that problems are more diverse and unpredictable than ever before, have led to a greater need for network management.

But even though the need is great, increased managability comes at a cost. Networked devices such as switches come in both managed and unmanaged flavors, and the managed ones are much more expensive. You must decide how much management capability you are willing to pay for based on the number of machines you will be connecting and the context and environment in which you will be connecting them. If you are connecting just three computers in your home, you will have little need for an expensive piece of hardware that has every management bell and whistle available. If something does go wrong, you can always reach over and unplug a problematic host. If instead you are connecting dozens of machines in a situation where you cannot simply remove one from the network by walking across the room or where you need be able to monitor the traffic levels of the connected devices, you probably will need at least some management capability from your network hardware. And obviously, if you are connecting hundreds or thousands of machines, you will have a definitive need for manageable devices so that you can ensure the stability of the network should a problem arise.

On the Massachusetts Institute of Technology (MIT) network, which has over 30,000 hosts connected to it, each host is attached to a managed network port. ^[1] As far as the field has advanced in the past few years, it is still the case that a single misbehaving host can cause a very large problem. Thus the ability to locate and potentially disable a problematic host is crucial.

^[1] This is a little bit of a lie. On all of the networks that are run by the Network Operations team, every host is on a managed port. There a few networks that are run by individual labs, and the labs may choose to do things differently.