The current Internet architecture nicely structures functionality into layers of protocols. While this reduces complexity, many tweaks have emerged because of the architecture's limited flexibility. Cross Layer Functionality corrodes the layer boundaries, intermediate layers had to be introduced for protocols like MPLS and IPsec, and middleboxes - like in case of NAT - further complicate the interaction of protocols. To overcome these problems, many publications have proposed modular solutions or protocol composition, allowing software engineering ideas to improve protocol design. Other publications state that instead of choosing a single common network architecture for the Future Internet, it might be advantageous to run multiple different architectures in parallel. We combine both approaches and make it possible to rapidly create and run different network architectures in parallel. While this allows for simplified Future Internet development, it requires the network architecture to be dynamically chosen. This paper not only presents a node architecture enabling the parallel operation of different network architectures but also introduces algorithms for their selection at runtime.

In this paper, we describe the approaches taken in the 4WARD project to address the challenges of the network of the future. Our main hypothesis is that the Future Internet must allow for the fast creation of diverse network designs and paradigms, and must also support their co-existence at run-time. We observe that a pure evolutionary path from the current Internet design will not be able to address, in a satisfactory manner, major issues like the handling of mobile users, information access and delivery, wide area sensor network applications, high management complexity, and malicious traffic that hamper network performance already today. Moreover, the Internet’s focus on interconnecting hosts and delivering bits has to be replaced by a more holistic vision of a network of information and content. This is a natural evolution of scope requiring nonetheless a re-design of the architecture. We describe how 4WARD directs research on network virtualisation, novel InNetworkManagement, a generic path concept, and an information centric approach, into a single framework for a diversified, but interoperable, network of the future.

In this paper, we present the application of machine learning techniques to the improvement of the congestion control of TCP in wired/wireless networks. TCP is sub-optimal in hybrid wired/wireless networks because it reacts in the same way to losses due to congestion and losses due to link errors. We thus propose to use machine learning techniques to build automatically a loss classifier from a database obtained by simulations of random network topologies. Several machine learning algorithms are compared for this task and the best method for this application turns out to be decision tree boosting. It outperforms ad hoc classifiers proposed in the networking literature.