Learning to detect and avoid run-time feature interactions inintelligent networks

摘要

The Intelligent Network (IN) allows rapid changes in the servicesnprovisioned and their functionality. Services may be supplied byndifferent service providers, making it unlikely that all servicenspecifications will be available for examination by any single agency.nApproaches to handle feature interaction problems must be able tonoperate within these constraints. Work by the authors has produced angeneric run time feature interaction manager (FIM) concept to managenfeature interactions in a live network. It monitors features as blacknboxes, learns their “correct” behavior and uses this tondetermine when feature interactions have occurred. The paper describesnand compares experiences using three different techniques to realize thenproposed approach. These are: state sequence monitoring, artificialnneural networks (ANN), and rule based monitoring which also includesnintegrated generic resolution approaches. The paper explores the designnalternatives with the various techniques, and reports on the resultsnobtained from experimentation