Choosing to Manage your Network

At some level all network devices require some management. Even the most simple devices have physical management needs as they are commissioned and connected to a power supply. But most devices need some form of configuration to tell them what role they are to play in the network and precisely how to behave. Even when autoconfiguration protocols like the Dynamic Host Configuration Protocol (DHCP) are used to dynamically assign IP addresses and to download basic configuration information, a network operator will still want to use management operations to inspect the devices to discover what addresses they are using.

In practice, many network devices are complex, requiring a large number of configuration parameters. Many, if not most, of these parameters can usually use default values, but fine tuning may be necessary to ensure optimal functioning of the network, and that requires some form of management access to the device.

At the same time, it is crucial to the understanding of the operation of a network to be able to inspect each node and observe how it is behaving. What resources are active and how much traffic are they carrying? Who has provisioned those connections that are causing a bottleneck for the CEO's emails? Why can't I send any packets to that host? The background information needed to answer these types of question ranges from basic state information about the devices, through detailed data concerning the inner functioning of the devices and thorough statistics recording the number of errors, packets, and bytes.

In order to get the most meaning out of management information retrieved from a device, it is usually decomposed in a logical and modular fashion. So, for example, one might be able to access data about a whole router, the line cards on the router, the interfaces on each line card, the protocol components running on the router, and so on. Conversely, configuration is most flexible when it can be applied to the same logical components of the system.

A final management requirement is the ability to provision new services. This may require commissioning resources at each node along a path through the network, or if a signaling protocol is in use, simply issuing management requests to the starting point of the new connection.

So, at many levels it is impossible to operate a network without some form of management. The remainder of this tutorial introduces how to use standardized approaches to produce a coherent management strategy for the whole of the network, making it possible to debug the network more effectively and to reduce the management resources required to operate a network constructed from computers from different vendors.

Network management is an area in which most Internet service providers (ISPs) seem to struggle. The nature of their networks is constantly changing, and the market is continually driving them to provide new and different services. These changes put a strain on existing network management tools and require the ISPs to race to adapt their techniques to their customers' requirements. In previous years, managed Internet services were the highest requirement, but these days, enterprises are looking for their ISP to support intranet or extranet services. This means that the service provider needs to provide an entire "network" to an individual enterprise customer and not just a set of simple and unrelated connections to the Internet. The new network services are provided to the customer as virtual private networks (VPNs) across a common shared network infrastructure owned by the ISP. This sharing of network resources provides a new challenge to the network management capabilities of the service provider that must now be able to partition resources and share them between customers.