Divide and conquer in network management using event-driven network area agents

In network management, simple polling is still the most used method by central management systems for gathering information and checking the status of network nodes. This results in a lot of network bandwidth being used across the whole network for management tasks. The main cause of this problem is the popular centrally managed network paradigm that is used today, where a single network management system collects, aggregates and processes all data without the use of distributed subsystems or delegated tasks. We propose a system where management tasks can be distributed dynamically and delegated across the network in agents. Each of these agents can perform generic tasks or tasks specific to the segment on which it is located. Network management workload can therefore effectively be spread out over the agents and the manager, enabling disconnected problem handling while diminishing bandwidth usage by remote data reduction, thereby making management tasks more scalable.