Today, SNMP is the most popular protocol for managing diverse commercial, university, and research internetworks.

Standardization activity continues even as vendors develop and release state-of-the-art SNMP-based management applications. SNMP is a simple protocol, yet its feature set is sufficiently powerful to handle the difficult problems involved with the management of heterogeneous networks.

The organizational model for SNMP based network management includes four elements:

• Management station
• Management agent
• Management information base
• Network management protocol

The network management station (NMS) is usually a standalone workstation, but it may be implemented over several systems. It includes a collection of software called the network management application (NMA). The NMA includes a user interface to allow authorized network managers to manage the network. It responds to user commands and issued commands to management agents throughout the network. The management agents are key network platforms and devices, other hosts, routers, bridges and hubs, equipped with SNMP so that they can be managed. They respond to requests for information and requests for actions from the NMS, such as polling, and may provide the NMS with important but unsolicited information, such as traps. All the management information of a particular agent is stored in the management information base on that agent. An agent might keep track of the following:

• Number and state of its virtual circuits
• Number of certain kinds of error messages received
• Number of bytes and packets in and out of the device
• Maximum output queue length, for routers and other internetworking devices
• Broadcast messages sent and received
• Network interfaces going down and coming up

The NMS performs a monitoring function by retrieving the values from the MIB. The NMS can cause an action to take place at an agent. The communication between the manager and the agent is carried out by an application layer network management protocol. SNMP uses User Datagram Protocol (UDP) and communicates over ports 161 and 162. It is based on an exchange of messages. There are three common message types:

• Get – Enables the management station to retrieve the value of MIB objects from the agent.
• Set – Enables the management station to set the value of MIB objects at the agent.
• Trap – Enables the agent to notify the management station of significant events.

This model is referred to as a two-tier model. However, it assumes that all network elements are manageable by SNMP. This is not always the case, as some devices have a proprietary management interface. In these cases, a three-tiered model is required. A network manager who wants to obtain information or control this proprietary node communicates with a proxy agent. The proxy agent then translates the manager’s SNMP request into a form appropriate to the target system and uses whatever proprietary management protocol is appropriate to communicate with the target system. Responses from the target to the proxy are translated into SNMP messages and communicated back to the manager.

Network management applications often offload some network management functionality to a remote monitor (RMON) probe. The RMON probe gathers management information locally, and then the network manager periodically retrieves a summary of this data.

The NMS is an ordinary workstation, running a typical operating system. It has a large amount of RAM, to hold all the management applications running at the same time. The manager runs a typical network protocol stack, such as TCP/IP. The network management applications rely on the host operating system, and on the communication architecture. Examples of network management applications are Ciscoworks2000, HP Openview, and SNMPv2c.

As discussed before, the manager may be a standalone, centralized workstation sending out queries to all agents, no matter where they are located. In a distributed network, a decentralized architecture is more appropriate, with local NMS at each site. These distributed NMS can act in a client-server architecture, in which one NMS acts as a master server, and the others are clients. The clients send their data to the master server for centralized storage. An alternative is that all distributed NMSs have equal responsibility, each with their own manager databases, so the management information is distributed over the peer NMSs.

Web Links SNMP and MIB Basics http://www.adventnet.com/products/ cagent/help/ Preface/ intro_snmpmib.html