Tuesday July 16, 2019
Thesis for the Degree of Habilitation
Queueing models have been a solid foundation for conceptualizing and dimensioning communication systems since the beginning of the 20th century. Current efforts to design Future Internet protocols and mechanisms require deep insights of the interaction of the communication systems and the applications exchanging data on top of them. Classical queueing theory, as well as, the emerging theory of network calculus provide insights into Quality of Service (QoS) questions encountered in the analysis of single and networked communication systems. Such questions comprise fundamental properties of packetswitching, data traffic scheduling, as well as, router buffer dimensioning.
In the last decade, modern communication networks have seen a radical change in architecture and requirements. On the one hand, the emergence of the so-called Internet of Things, i.e., the endowment of everyday physical and virtual objects with Internet connectivity, did not only lead to a tremendous growth in data traffic but also to much stricter requirements for QoS guarantees, e.g., for delays in industrial applications. On the other hand, the rise of Software-Defined Networking and Network Function Virtualization as concepts and tools to reconfigure the functionalities of communication networks provided a wide range of possibilities to control communication systems at runtime. These changes in network architecture entail new and strict Quality of Service requirements but they also enable a wide range of adaptive approaches to communication network management and resource utilization. The goal of this work is to review central contributions on the foundations and applications of adaptive queueing systems in communication networking. This work particularly highlights approaches and contributions by the author to the aforementioned area.
While multiple comprehensive expositions exist that illustrate the application of queueing theory and network calculus to the analysis and composition of elements of communication systems, our focus lies in this work on approaches to adaptivity as motivated by the recent development of network architectures. This work provides an overview of approaches to adaptive resource management in communication systems that are based on queueing models of the communication network elements. The reviewed works show that substantial resource savings can be achieved through adaptation as adaptive approaches make a more efficient use of communication resources than static counterparts. This work also shows that introducing adaptivity in parallel server systems helps to retain Quality of Service guarantees in face of dynamic conditions of the communication environments.
Notable applications of the models of adaptive queueing systems that are treated in this work in the context of communication systems are illustrated in Fig. 1.1. These applications span the core and the edge of modern communication networks including queue-aware transmission resource allocation in wireless systems, collaborative wireless transmission, adaptive scheduling in parallel data processing, application-aware multipath communication and network-assisted and queue-based adaptive bitrate video streaming.