Automatic event-driven system for network management

It is very tedious and trivial if a network administrator has to manage all the equipment manually. Unfortunately, there are usually numerous and heterogeneous devices over which an administrator must watch. This paper proposes the design and development of an automatic event-driven system to reduce the burden on network administrators. By allowing managers to pre-define the action of some conditions, the network management system can quickly respond to critical events and mitigate the network trouble, in contrary to many commercial network management systems which can only send a notification to administrators. For example, if a user tries to download 10 GBytes of files in a single day, which is a behaviour violating the acceptable user policy (AUP) in campus, it may be necessary to limit the bandwidth of the user in order to prevent the campus network from being jammed because of a single user. With our system, an administrator can easily configure the network management system to automatically take actions so that the managers need to do nothing to impose appropriate penalty on individual users. The system will control devices by Simple Network Management Protocol (SNMP), which is supported by most network devices such as routers and switches. In this paper, we implemented the system in our campus, and chose the switches as the device to control user behaviours. This system obtains information from devices by SYSLOG, while it sends instructions to devices through SNMP. In a real case we found that when a device breaks down, the system can solve the problem automatically and immediately take responses as defined by the manager. The manager will get the report by emails and MSN messages, too. In case of equipment upgrades where some devices will stop working temporarily, the manager can also modify the reaction through this system to prevent it from sending out false alarms. This system provides managers a convenient tool to handle abnormal situations in real-time monitoring.