Adaptive QoS-aware resource allocation for wireless networks

Sammanfattning: Wirelesscommunication networks are facing a paradigm shift. From providingonly voice communication, new generations of wireless networks aredesigned to provide different types of multimedia communicationstogether with different types of data services and aim toseamlessly integrate in the big Internet infrastructure. Some of these applications and services have strong resourcerequirements in order to function properly (e.g. videoconferences),others are flexible enough to adapt to whatever is available (e.g.FTP). Also, different services (or different users), might havedifferent importance levels, and should be treated accordingly.Providing resource assurance and differentiation is often referredto as quality of service (QoS). Moreover, due to the constrainedand fluctuating bandwidth of the wireless link, and user mobility,wireless networks represent a class of distributed systems with ahigher degree of unpredictability and dynamic change as compared totheir wireline counterparts. In this thesis we study how novel resource allocation algorithmscan improve the behaviour (the offered QoS) of dynamicunpredictable distributed systems, such as a wireless network,during periods of overload. This work concerns both low levelenforcement mechanisms and high-level policy dependent optimisationalgorithms. First, we propose and evaluate adaptive admission controlalgorithms for controlling the load on a processor in a radionetwork controller. We use feedback mechanisms inspired byautomatic control techniques to prevent CPU overload, andpolicy-dependent deterministic algorithms to provide servicedifferentiation. Second, we propose and evaluate a QoS-aware bandwidth admissioncontrol and allocation algorithm for the radio link in a networkcell. The acceptable quality levels for a connection are specifiedusing bandwidth dependent utility functions, and our scheme aims tomaximise system-wide utility. The novelty in our approach is thatwe take into account bandwidth reallocation, which arise as aconsequence of the dynamic environment, and their effects on theaccumulated utility of the differentconnections.