Softwarized Adaptive Control of Network Monitoring Systems

Softwarization and programmability allow for flexible and more efficient network resource utilization. This leads to more dynamic in traffic and behavioral patterns and therefore demands for comprehensive network monitoring. Furthermore, a sound understanding of the network, its resources, and the traffic carried is required for being able to conduct sensible control decisions. This comes along with monitoring data analytics, which is challenging from a resources perspective. The AutoMon project [1] works on a monitoring solution using closed loop control for gaining maximum insight at minimal resource utilization. We developed a control concept, where a controller dynamically adapts the monitoring functions in the network as well as the data analytics part of the network monitoring system. While the concept also covers component metrics and active measurements, our current focus is on flow monitoring as it is most challenging due to the high and often unpredictable amount of data. Our first, yet simple, control algorithms focus on the balancing of storage consumption, as this is the most critical resource. Evaluations by simulation show that they are feasible in clusters with heterogeneous resources and typical flow rate patterns. We implemented a FlowMediator component for dynamic distribution of flow monitoring data, which can take control commands from a controller running these algorithms. Although, we use Software Defined Networking (SDN) concepts, SDN is no prerequisite. Hence, our approach is suitable for brown-field deployments and has been validated in labs using traffic from a large production network.