Lucent Technologies (former AT&T) developped Total Network Management (TNM) a very large product which facilitates early problem detection and prompt repair of telecommunication network facilities and switches. Network alarms are analysed to build up an overall picture of the network's current failures.

Major telecommunication hubs employ a number of operators to control maintenance of the telephone network. Due to the volume of alarm messages, (approx. 15 000 per hour), operators were unable to keep pace with the flow of information across their screens. An Event Correlation Expert feature package, ECXpert has been realised and incorporated into the TNM product family. ECXpert semi-automates the interpretation of alarms and enables operators to concentrate on crucial network failures.

ECXpert models the real-world situation by means of event correlation. As events (network failures) arrive, a tree is built-up which represents the dependency between events. Certain events can be determined to be caused as a direct result of other events which represent the true cause of the network failure. By correlating the events, the key problems can be isolated and displayed at higher priority for the operators. Repair crews are then sent to repair the faults faster and before the problem escalates.

ECXpert combines C++ and IF/Prolog code to provide support for object-oriented design with Prolog's reasoning and meta-language manipulation. TNM customers specify rules to co-ordinate their specific telecommunication networks needs; these are then compiled and acted upon by ECXpert together with live events. The event-simulation and correlation provides a high-level view on network failures.

TNM is in use 24hrs per day, every day; IF/Prolog's advanced garbage collection technology make it suitable for such real-time continuous operation. Stability is very important, TNM makes use of AT&T's own fault-tolerant Starserver machines.

The system is currently installed in several major US telecommunication firms and is being marketed on a world-wide basis. It is estimated to save over $500 000 annually, in downtime and maintenance costs, at each node of its operation.

Hardware : AT&T Starserver UNIX mainframes with attached terminals.

Software : IF/Prolog, C++.