Traffic Studies for DSA Policies in a Simple Cellular Context with Packet Services Hany Kamal, Marceau Coupechoux, Philippe Godlewski TELECOM ParisTech & CNRS LTCI 46, rue Barrault, Paris, France {firstname.lastname}@Telecom-ParisTech.fr Abstract—DSA (Dynamic Spectrum Allocation) techniques are very challenging when the quality of service has to be guaranteed in a flexible spectrum situation. In this paper, we present and analyze DSA policies for packet services in cellular context. A centralized model, where a meta-operator shares a common spectrum among different operators, is considered. We focus on two criteria for the policies design: the total welfare (sum of operators’ rewards), and the blocking probability. We go through two steps to pass from the actual FSA (Fixed Spectrum Allocation) situation into DSA. First, DSA algorithms depend on the arrival rates. Second, DSA algorithms depend on both the arrival rates as well as the number of active users. Targeting the reward maximization shows to be inefficient when the blocking probability has to be guaranteed. However policies targeting a blocking probability threshold, achieve greater rewards then FSA rewards. We also present a heuristic DSA algorithm that takes into consideration: the arrival rates, the number of active users and the blocking probability. The algorithm gives a very close blocking probability to the one achieved using FSA, while the obtained reward significantly exceeds the FSA reward. I. I NTRODUCTION The actual spectrum crowd situation and the rapid evo- lutions of the SDR (Software Defined Radio) techniques, provoke the development of cognitive radio and DSA systems. Existent spectrum allocation process, denoted as FSA (Fixed Spectrum Allocation), is inflexible and shown to be inefficient [1]. In [2], the spectrum management models are divided into four main axis: command and control, exclusive-use, primary/secondary usage, and commons. The exclusive-use model presents the actual cellular operator case where the operators own exclusively the rights to use the spectrum band for decades. DSA algorithms are being investigated as new promising techniques to overcome this inflexible situation which has leaded to resources limitation problem. In the cellular context two main axis of resource manage- ment exist. The JRRM (Joint Radio Resource Management) axe, in which one operator manages jointly his networks (or Radio Access Technology) making benefit of his own licensed bands [10]. The second axe: operator sharing DSA, (or Inter-operator DSA) in which the competition and/or the cooperation aspects between different operators are explored. Competition aspects are referred to the costs and revenues par- titioned among the operators as a result of spectrum sharing. Several researchers have worked on DSA for cellular net- works. For instant in reference [3], the authors introduce an inter-vendor spectrum sharing technique where a spectrum allocation server dynamically assigns the unused spectrum from a vendor to another. The authors in [7], made use of a Darwinian algorithm for DSA in WCDMA networks, the algorithm has been proposed to assign the minimum number of carriers to the cells while satisfying the required demands. In [6], authors propose a coordinated DSA system where a pool of resources (CAB ”Coordinated Access Band”) is shared and controlled by a central entity (the regional spectrum broker). Most of the works related to DSA in cellular networks are based on centralized architecture, due to its practical impact [4]. We believe the success of a centralized DSA network model depends on to the proper design of the centralized entity (meta-operator). In this paper we focus on the DSA algorithms for the cellular networks in a spectrum sharing context. We present and evaluate DSA strategies (policies) for the meta- operator. Talking about sharing the spectrum argues the consideration of the pricing aspects. As cellular operators pay high prices for the license, hence their main interest in sharing the spec- trum lies behind the expected benefits [8]. Many references considered the pricing issue while studying DSA techniques. In [9], the authors analyze a network model where the base stations of the service providers are sharing a common amount of spectrum. A distributed DSA algorithm is proposed where each user maximizes his utility (bit rate) minus the payment for the spectrum. The revenue maximization in a spectrum auction framework has been studied also in [5]. In [4], the authors propose a dynamic auction approach to allocate the spectrum to competing base stations. The base stations are sharing a common spectrum band controlled by a spectrum broker. The broker assigns the spectrum to the base stations to maximize its revenue, without violating the interference constraint. Most of the works dealed with DSA in cellular networks have supposed the demands (in terms of spectrum) are already given [7], [3], [6], and [4]. We are focusing in this paper on the traffic and dimensioning study. The target of our study is to determine how much spectrum the operators need to buy. Authors in [9] have performed a similar study, however (unlike this paper) they presented a distributed algorithm for the end- users. The authors in [9] did not consider the impact on the blocking probability. We present a network model where different operators are