An information-theoretic approach of network structure and dynamics

Abstract

Modern networks are facing the most challenges in recent years. Life is changing, and the Internet became one of the essential services such as power, healthcare, banking. The Internet has opened global information, and nowadays, a stable and efficient network is a global right and requirement. Legacy networks inherited many concepts and infrastructure from circuit-switching networks, with inflexible design relying most on hardware. Another critical aspect of networking is debugging it, which requires specialized personnel with properly designed software. For these reasons – and currently increasing demand – Internet Service Providers are dealing with challenges that require innovation and scientific strategies. This scenario led to new paradigms such as software-defined networks (SDNs) to address most of the current issues by bringing the network logic from hardware design to software development and virtualization. The emergence of SDNs has drawn the attention of network scientists and engineers to new roads. The Internet is the most extensive distributed system, and it relies upon distributed devices communicating efficiently through protocols. A computer network is a sophisticated collection of network devices and end-systems. SDNs allow a protocolagnostic, centralized view, and flexible control of the network, favoring the development of new strategies that are not achievable into legacy IP networks. In this work, we study two main aspects of computer networks: structure and dynamics. By understanding the network characteristics and modeling its dynamical processes, we can uncover how network structure and dynamics affect its robustness. To achieve this understanding, we propose the usage of information-theory quantifiers for network characterization. For network topology, we introduce the Fisher Information Measure for quantifying the network characteristics, using alongside the Network Entropy in a bi-dimensional representation that allows us to identify if a network is closer to a random, small-world or scale-free topology.