Monitoring and Observation for Large Scale Complex Systems Using Network Graphs

The pervasiveness and ubiquity of computing systems has seen computational devices and methods become accepted as a part of everyday life. This has brought forward wide area, complex and highly distributed computing environments, which become prone to poor performance or failure as the system evolves and adapts to its users. To more efficiently manage these systems a monitoring system is required that provides both timely and accurate data regarding the state of the system. The scale of the systems, however, precludes any full-scale instrumentation for control by human or centralised management system. It is, therefore, proposed here to use techniques from network graphs whereby distributed monitoring can be achieved through the intelligent selection of monitoring points to give as near as possible full-scale observation. We present a formal representation of a fully distributed monitoring technique, based on a network of monitoring points (nodes) and report on its initial use in a simulation environment.