| The proliferation of computer networking technologies have enabled many emerging paradigms such as mobile social networking (MSN), smart grid, and Internet-of-things (IoT). In these paradigms, the computer networks not only support, but also tightly interact with other cyber (i.e., social networks) and/or physical (i.e., power and infrastructure networks) entities, forming an interdependent cyber and physical networks. Unfortunately, the interdependence relation in cyber-physical networks poses a major challenge in analyzing the networks' performance. Moreover, various man-made and natural threats may induce faults, which due to the interdependence can spread to a larger part of the cyber-physical networks, causing a so-called cascade-of-failures. Since billions of people will depend on cyber-physical networks, performance degradation-especially under threat-induced cascade-of-failures|will have a profound impact and remains as an open yet interesting problem.;In this dissertation, we aim to understand the performance of cyber-physical networks and its robustness against man-made and natural faults. Because it is not practical to provide a unified framework for all types of cyber-physical networks, which have their own features and objectives, we adopt a systematic, step-by-step approach to model and evaluate the networks' performance and robustness. Specifically, we focus on four emerging paradigms: D2D-based IoT, IoT-applied dynamic spectrum access (DSA) network, D2D-based MSN, and smart grid.;In particular, we first consider the intra-network interdependence in the communication network by studying the delay of reaching a point-of-access in D2D-based IoT applications. We propose a packet mobility model to capture the progress of data packets toward the point-ofaccess, which is used to derive the upper and lower bounds of access delay. Then, we propose a packet shedding scheme that reduces the total transmit power while guaranteeing that packets can be delivered within a pre-determined access delay. Second, we study how to quickly establish a common control channel between a pair of DSA nodes with limited channel hopping capability. We introduce graph-based channel hopping schemes with quick rendezvous and show that the proposed schemes achieve satisfactory rendezvous rate under an indoor mobile environment. In addition, we consider the inter-network interdependence between the cyber and physical networks by analyzing the impact of initial node dropouts to the network-wide and end-users' resilience of D2D-based MSNs. Finally, we characterize the spatial and temporal impacts of edge disconnections in generic cyber-physical networks. The works in this dissertation not only expand our knowledge on the performance of cyber-physical networks, but also provide instrumental guidelines to the design of robust interdependent networks. |
