Skip to Main Content
Enterprises have become increasingly dependent on information technology capabilities (e.g. secure remote access for mobile users) to support their business objectives. Consequently, determining which users are affected by component failures remains a very important and challenging problem. Analyzing operational impact requires an understanding of how the system components are inter-dependent, and when the components are actually employed by the system users. Our approach collects monitoring data from the end systems. Data mining and analysis are used to infer system dependency topologies and usage patterns. We compare centralized, partially distributed, and fully distributed implementation approaches using computers connected to a campus-wide system. The results show that distributed approaches can be used to minimize the amount of data transmitted between systems, without significantly reducing the overall quality of the impact analysis. These distributed approaches will support efficient and scalable impact assessment in modern enterprise systems.