Designing for the Future: A Complex Systems Approach to Communication Networks
Citation:DZAFERAGIC, MERIM, Designing for the Future: A Complex Systems Approach to Communication Networks, Trinity College Dublin.School of Engineering, 2020
PhD_thesis_Merim_Dzaferagic.pdf (PDF) 4.370Mb
The network size and the deployment density of wireless networks continue to increase from year to year. Additionally, networks are experiencing an operational shift that affects both: (1) the network architecture, and (2) the service implementation. The network architecture is changing from a traditionally rigid hierarchical - hardware first - to a more flat and flexible - software first - implementation. This shift enables innovation related to, among other things, network infrastructure ownership, dynamic resource sharing, on-demand resource allocation. The evolution of the network architecture underpins the shift related to the service implementation. New services have strict requirements (e.g. latency, throughput, reliability), dynamically demanding resources on a more granular level. The growing size and operational changes demand scalability and adaptability to be part of the network design. Not only are individual networks becoming more sophisticated, but it is increasingly infeasible to consider any one kind of network in isolation; networks such as cellular, Wi-Fi, vehicular and IoT, increasingly have interdependencies. In addition, many subsets of networks are no longer centrally planned and rolled-out by a single owner, but evolve over time based on user deployed infrastructure. The network can be viewed as a living organism, that evolves over time and adapts to the changes in its environment. Focusing on making nodes more capable and intelligent as done to date by the cognitive radio community often results in higher cost, limiting the scalability of the adopted techniques. In this thesis, we propose a complex systems science approach to communication networks. We focus on three complex systems principles, i.e. complexity, degeneracy and emergence. Additionally, we divide the complex systems tools and techniques into three layers, i.e. analysis, modeling and design. The problem that we address in this thesis can be stated in the form of the following research question: "What does the analysis of the micro-scale structures tell us about the macro-scale performance, and how do the local interaction rules lead to global organization/synchronization in communication networks?" Our first step was to identify what aspects of the network (e.g. infrastructure, network functions, signal processing) could be analyzed in the context of complex systems. The focus is on network functions (e.g. clustering in WSN, frequency allocation in cellular networks, allowing us to have a better understanding of the impact that different protocol procedures have on the network operation. The research question can be broken down into three parts, which drove the development of the thesis: 1. How to quantify the impact that micro-scale structures have on the macro-scale performance of communication networks? 2. How to identify macro-scale/system-level topologies that are constructed out of diverse micro-scale structures (i.e. degenerate structures)? 3. How to achieve global organization through limited knowledge and local interactions? We start with a discussion of different tools and techniques that have been developed and applied to understand unexplored system properties (e.g. the capability of a system to store, communicate and process information) in sciences like physics, neurobiology, urbanism, social networks, etc. We address the above-mentioned questions by resorting to a wide range of tools for analysis (e.g. network science, information theory, statistical mechanics), modeling (e.g. Equation-Based Modeling, Agent-Based Modeling) and design (e.g. Reference Design, Trial and Error, Analytical Approach). We also discuss the evolution of communication networks, showing that the need for flexibility, scalability and adaptability demands new approaches to analysis, modeling and design. We then propose a framework to model the underlying structure of network functions with graphs, allowing us to study the organizational characteristics of network functions as complex systems. In order to quantify the impact that micro-scale structures have on the macro-scale performance, we propose a complexity metric called functional complexity. This allows us to correlate the organizational structure to the performance of different network functions (e.g. frequency allocation in cellular networks, clustering in WSN). Then we shift our focus from complexity to degeneracy and reapply the same modeling approach, i.e. modeling network functions with graphs. Here we focus on the identification of macro-scale/system-level topologies that are constructed out of diverse micro-scale structures in IoT networks by studying degeneracy, i.e. the multiplicity of computational graphs that allow us to perform the same computation by using different subgraph structures. Finally, we move on to the third layer, i.e. design. We design local rules of interaction between base stations in a cellular network that emerge in a desirable global property, i.e. formation of handover regions that minimize the signaling and latency related to mobility management. Here, we apply the tools from all three layers, i.e. analysis, modeling and design, showing the full potential and benefits of the application of complex systems tools to design scalable, adaptive and robust network functions for future communication networks.
Author: DZAFERAGIC, MERIM
Publisher:Trinity College Dublin. School of Engineering. Discipline of Electronic & Elect. Engineering
Type of material:Thesis
Availability:Full text available