Fair Scheduling and Resource Allocation in Packet Based Radio Access Networks

Abstract: Wireless communication systems are by nature and by regulations limited in bandwidth and are therefore limited in capacity. Services with Quality of Service (QoS) requirements will enforce a more efficient utilisation of the available capacity. This observation motivates the development of schedulers that can deliver the required QoS as well as being resource efficient. In this thesis, we study resource allocation schemes and scheduling algorithms in packet radio networks. The aim of this thesis is to develop resource efficient allocation schemes capable of keeping deadline requirements for real-time services as well as providing minimum bandwidth for best-effort services. The first part of the thesis focuses on scheduling traffic in Mobile Broadband Systems (MBS). MBS is an extension of the wireline Broadband Integrated Services Digital Networks system providing radio coverage restricted to a small area as well as supporting mobility. The proposed scheduling algorithm is based on Weighted Round Robin, but the method for assigning the weights to the different traffic streams is different from the widely accepted. The proposed scheme allows the user and network operator to define their own priorities and own methods for differentiating the traffic streams. The second part of the thesis focuses on resource allocation and scheduling in the HiperAccess radio access network. Delay bounds for deadline driven services are presented using the method of Network Calculus (NC). Unfortunately, due to the functionality of HiperAccess and the nature of NC, these delay bounds are very pessimistic and therefore not very resource efficient. This has motivated the development of decision methods which will aid the resource allocation in making an efficient allocation of the uplink capacity. The proposed methods, which rely on the ability of predicting the terminals' future need for bandwidth, significantly increase the system utilisation with very few deadline misses. Further, a new definition of fairness and a method for calculating the fairness of a given traffic mix is presented. The proposed notion of fairness is better suited for time-slotted systems with service vacations carrying different traffic types, than the commonly used definition of fairness.