Abstract
The reliance on computer communications networks for business and commerce, education, entertainment and many other applications demands these resources are managed effectively. In this context, “management” refers to ensuring security and performance, recovering from faults and accounting for utilisation. Each of these activities requires knowledge of the network configuration, and information about the networked devices. It is on the basis of this knowledge that management decisions to change the configuration and network behaviour are taken. Hence the manager requires as full and accurate a set of information about the network under their control as it is possible to get. Typically, this information resides across the network and is transferred (to the network manager) to assist the network manager’s decision making.
TopNetwork Management: An Overview
The history of network management in the form discussed here starts in the late 1970s, with standard protocols to monitor network devices, first gateway devices and later network host systems and interconnection devices. It is important to realise that in typical computer communications networks (e.g., a TCP/IP based system), the same physical network carries both user and management data. This differs from the telecommunications approach of a separate “management network.” Hence, in computer networks, managers and users share and compete for bandwidth, processing, and storage capacity.
Client-Server Based Network Management
In its simplest, in a deployment model exemplified by SNMPv1, a single, central network management platform receives information from network devices by “agent” programs resident on the devices in response to management platform requests. The data are then collated and presented to the (human) network manager (Figure 1). This simple client-server, single point of concentration approach creates a single point of failure or loading, since more management data are generated in emergencies, possibly making the original problem even worse.
Figure 1. SNMP in a centralised system
The Search for Alternatives
Concerns over data volume and concentration have led researchers to explore alternatives. In particular, addressing the problem of the quantity of data transferred created improvements in the effective payload of SNMP, version 2’s bulk transfer operation among the most significant. Other proposals reduced the volume of data by preprocessing, introducing a hierarchy of managers so the required data need not travel so far with decisions taken nearer the point of impact.
Key Terms in this Chapter
Management Agent: A software process able to support some or all of the network management activity required on a networked device. Agents are able to access and, where appropriate, modify data on the devices themselves. Agents are of two major flavors: static agents, which reside on the managed device and respond to management requests; and mobile agents, which are able to transfer between devices, collecting data, making decisions, and affecting device behavior as they move.
Power-Saving Network Management: A growing awareness that networked devices are a significant power drain, even when inactive. The traditional network management processes (device polling, health checks, fault and intrusion detection by anomaly) are predicated on the requirement that networked devices are able and ready to report their status (and to monitor the status of neighbors) on a regular basis, and where a lack of response is taken as an indication of potential device or connection failure. Therefore, networked devices typically remained powered on so that they can respond to checks on a regular basis.
Centralized/Distributed Management: In small-scale networks, it is possible to create a centralized network management platform, to which all management data is delivered, at which all decisions are taken, and from which all management commands are issued. There are clearly limitations to the size of the network which can be controlled in this manner, limits being created by the bottlenecks of transfer and processing of data at a single location.
Network Management: The process of controlling a computer network to ensure it delivers an appropriate service to its users. This term also refers to the software and other equipment used to support this activity. Much of the information required by a network manager is located in the devices which make up the network, in particular, monitoring stations; and producing a complete understanding of the network behavior requires information from different locations, hence either the decision maker must fetch the data, or the decision maker must travel to the location(s) of the data.
SNMP: The Simple Network Management Protocol is an IETF communications protocol designed to support the function of network management. In its most basic form, it provides support for data retrieval, for modifying data in controlled devices, and for generating alerts. Extensions allow communication between network managers. The original SNMP, now called SNMP version 1 (SNMPv1) is a simple client-server protocol, with no in-built security; successive enhancements brought about SNMPv2, which included more efficient methods for handling larger volumes of data, and SNMPv3, with in-built authentication, encryption, and other security features.
XML-Based Network Management: A proposal to use the Extended Markup Language (XML) to support Web-based access to network management facilities in place of an SNMP-based mechanism. This is expected to have the advantage of making use of universally available Web support software, so removing the need for explicit SNMP protocols, therefore enabling the network management toolset to be richer and more easily provided. A possible disadvantage is the extra load on the network created by this enhanced functionality.
MIB: The Management Information Base is a virtual repository of network management data, conceptually located on the devices managed. Software called a management agent possesses the knowledge to extract real data necessary to populate the MIB, whether by direct retrieval of a specific data item or by some processing activity. SNMP requests and responses use MIB names to identify their data.