An algorithm for fault isolation and event correlation in integrated networks is presented. It reconstructs fault propagation during run-time by exploring relationships between managed objects and provides improved focus and efficiency compared to similar algorithms. The functionality of the algorithm is generic, straightforward, efficient, and applicable for different management architectures such as SNMP and TMN. Clearly defined interfaces allow parallel execution of problem investigation, integration of different management architectures and systems, and incorporation of other correlation techniques. Prototype and production level systems have been implemented for TMN- and SNMP-based management systems to validate the concept and for use in real networks. This development was done after an extensive study of requirements and existing techniques, the results of which are also presented.