Abstract
The implementation and evaluation of a production system shell which can negotiate a large amount of data over a database and is capable of responding in real time to external changes such as alarms, from networks is discussed with the goal of applying it to expert systems in such fields as network management. The matching algorithm is obtained by extending the LEAPS matching algorithm, based on the highest priority search, by incorporating external changes into it in real time. In addition, data are stored persistently in a database and such persistent data can be used for inference by differentiating them from the nonpersistent data of the main memory. The evaluation results showed that external changes can be reflected in inference in real time, that inference with respect to the persistent data can be realized with a performance 1/10 to 1/50 of that in the case of nonpersistent data by matching, and that it is possible to achieve highβspeed processing, several times to several tens of times as fast as past inference methods which store data in widely used database management systems (DBMS) and perform inference by accessing the data by SQL and the like from the expert system. Β© 2002 Scripta Technica, Syst Comp Jpn, 33(3): 11β20, 2002