Data Distribution and Control in Modern Networks

Data Distribution and control over the Internet and modern networks is a very hot issue these days. The aim of this thesis is to address, on one hand, data dissemination in terms of energy consumption over Content Delivery Networks (CDNs) and, on the other hand, network control and scalability issues regarding Software Defined Networking (SDN). Data dissemination via CDNs is a leading energy-consuming sector in the ICT area. A CDN system can be abstracted as a main server hosting the entire data set and several surrogate servers, each one housing a portion of the whole data set. In order to design the architecture of a CDN system and to plan network policies, having a clear idea about energy consumption of the network can play an important role. Thus, a realistic model is needed to compute energy consumption of a system according to network characteristics. This model can guide network architects to design an energy efficient network while keeping performance in the desired level. In this dissertation, a new model to compute energy consumption of CDNs is presented. Two major aspects distinguish this model from the existing ones. Firstly, the Internet topology considered in the model, despite existing models exploiting a flat graph, is a three-tier map which is close to the real Internet topology. The three-tier topology represents the three levels of ISPs forming the Internet. The second novelty of the model is that total energy consumption also comprises synchronization energy consumption, which is the energy consumed to keep all surrogate servers up to date and synchronized. In order to validate the model, a CDN simulator is designed to simulate the hierarchical network and to consider synchronization energy consumption. Results show the importance of synchronization and reveal how synchronization can affect the total energy consumption. Various caching strategies are compared and the impact of caching policies, content popularity, network topology and number of surrogates is discussed. In all mentioned scenarios, various content request and modification ranges are compared.