Chapter 9: Network Management

Overview

This chapter is concerned with the management and maintenance of internetworks. Management should be tackled as a proactive task and should form an integral part of your design strategy. Maintenance is all too often a reactive task, dealing with unforeseen problems and diagnosing network failures. Network management becomes a key concern as networks grow in both size and complexity. As the number of devices, services, users, and sites begins to multiply, so do the network manager's activities and responsibilities. In a large commercial internetwork, it may be critical to identify problems and resolve them in a matter of minutes. You need to identify and resolve problems before they become serious enough to jeopardize the business. Remote devices may need to be monitored, closed down, and restarted. Faults must be clearly reported, understood, and acted upon. Inefficiency, overloading, and degraded performance should be diagnosed and resolved proactively. To achieve all this, management centers and managed devices must be capable of communicating with one another. These functions are commonly referred to as a Management Information Services (MIS). If the network is designed properly from the ground up, the MIS should be used as a proactive tool to monitor the network and work out where performance can be improved and cost savings made. Value-added services (such as advanced network monitoring and modeling tools) can be built into this basic structure. In a badly designed network you will spend the majority of your time fire fighting.

Perhaps the two classic vendor architectures are DEC's DECnet and IBM's SNA. Both architectures have strong emphasis on management and introduce very comprehensive and well-integrated management models. In the IP world SNMP has been around since the late 1980s, but since no single vendor controls this architecture the management model is less comprehensive. Nevertheless, from an internetwork point of view SNMP currently dominates the installed base, and even proprietary systems often have gateways into the SNMP domain.

Network management and control represents a potentially exciting area for further investment in the application of artificial intelligence. Networks are destined to evolve from disjoint, inarticulate, and helpless organisms to intelligent, self-maintaining entities capable of keeping themselves in peak performance with minimal intervention from the network manager. Serious network outages are regularly caused by simple cable breaks, loose connectors, faulty hardware, and misconfigured software. These problems could be better resolved through good design choices and proactive management tools (especially tools that distribute intelligence across the network); this represents a promising area for further research.