Performance Analysis of Scheduling Policies for Delay-Tolerant Applications in Centralized Wireless Networks Mohamed Shaqfeh and Norbert Goertz Institute for Digital Communications Joint Research Institute for Signal & Image Processing School of Engineering and Electronics The University of Edinburgh Mayfield Rd., Edinburgh EH9 3JL, Scotland, UK Email: {M.Shaqfeh, Norbert.Goertz}@ed.ac.uk Abstract— Many scheduling schemes have been proposed in literature to control how different users access a wireless channel. Channel-aware schedulers exploit the measurements of instan- taneous channel conditions of the different users to obtain throughput gains by proper selection of the users to transmit (or receive) in each channel state. In this paper, the performance of channel-aware scheduling policies which are applicable for delay-tolerant applications in centralized wireless networks are analyzed in a new mathematical framework. The framework is applied in numerical examples to compare the performance of different scheduling policies in terms of their efficiency in allocating the wireless resources. This efficiency is measured against the set of all possible operating points of the system, chosen, e.g., by the network operator. Index Terms – centralized wireless networks, channel-aware scheduling, power allocation, delay-tolerant applications I. I NTRODUCTION An important feature of multiuser communication over fading channels is multiuser diversity: The independent fading characteristics of the users’ channels increase the probability of having one or more users with very good channel condition at each time instance. The multiuser diversity gain can be obtained by exploiting the independent fading conditions, and scheduling only the users with good channels. During the last decade, the design of multiuser scheduling schemes for wireless networks has been extensively studied. The algorithms which were state-of-the-art few years ago [1], are now outperformed by a lot of new schemes. Proposals in literature differ in their policies to choose the scheduled user based on channel conditions. The objective of this work is to systematically compare, in a new universal framework, known channel-aware scheduling algorithms that are applicable for delay-tolerant applications in centralized wireless networks. Algorithms that consider strict delay-constraints are, however, not included in our analysis. In centralized networks, the scheduling decisions for both the uplink and the downlink are taken at the wireless ac- cess point or the base station of the network. This unit is provided with a rich set of information such as the traffic load and the different QoS 1 requirements of the served traffic classes as well as the instantaneous channel conditions of the wireless links to active users. We assume in our analysis that the channel variations are not too fast so that the effect of the channel measurement delay [2] is negligible and the channel coefficients can be estimated at the receiver and be communicated to the transmitter with sufficient accuracy at low overhead. This also means that we can perform coherent detection at the receiver, i.e., we get no phase error and the “I” and the “Q” components are both scaled by the magnitude of the channel’s fading coefficient which is the square root of the channel power gain. For our analysis we assume general schedulers which may, e.g., be applied in OFDMA 2 systems with dynamic sub-carrier allocation (DSA) and adaptive modulation and coding (AMC). The OFDMA scheme enables the exploitation of multiuser diversity in the frequency domain (due to frequency-selective channels in the total transmission bandwidth) and in the time domain [3]. Therefore, OFDMA provides a rather general setting with other (e.g. single carrier) systems as special cases. The scheduling decisions, i.e. the decisions which of the users are allowed to access the channel, are taken for each channel block over which the channel fading coefficients can be considered to be constant. Each channel block may consist of many slots (in time or frequency or both) over which the channel fading coefficients do not change as the slots belong to one channel block. Some of the scheduling algorithms we include in a case study to demonstrate our new analysis were originally pro- posed and applied for single-carrier systems (e.g. the propor- 1 Quality of Service 2 Orthogonal Frequency Division Multiple Access SPECTS 2008 309 ISBN: 1-56555-320-9 Authorized licensed use limited to: Universitatsbibliothek der TU Wien. Downloaded on May 29, 2009 at 04:19 from IEEE Xplore. Restrictions apply.