Cognitive management of procotol composition enhanced Future Internet elements

This paper presents the final architecture for Cognitive Future Internet Element Management and in-network Dynamic Protocol Composition, as well as the results and insights gained by an prototype implementation and experiments performed. The high complexity of today's network infrastructures and the demand for reliable and robust but inexpensive networks stimulated research on evolving networks to become more autonomous and therefore reducing the need for human intervention. The presented approach is based on integrating a Monitoring, Decision-Making, and Execution (MDE) cycle into the situation aware Network Element Cognitive Manager (NECM). Accepting that the end-to-end paradigm does not hold for all mechanisms, we enable the NECM to control in-network functionality. For the prototype the NECM is complemented by the Dynamic Protocol Composition Framework (DPCF) which allows to control and execute arbitrary protocol functionality in intermediate nodes. For the experime nts we considered a wireless mesh backhaul network which is likely to suffer from packet loss depending on external conditions like weather, reflections on moving objects etc. We could show that packet loss or overload have been recognized and appropriate actions have been taken. The analysis of the experiments showed that the different phases of the M-D-E cyle differ in time consumption by several orders of magnitude which requires architectures for cognitive network management to be designed to align the different processes to achieve optimal results.