Simulative Analysis of a Multi-cell Admission Control Algorithm in WCDMA Networks Gábor Fodor Ericsson Research SE-164 80 Stockholm, Sweden Gabor.Fodor@ericsson.com Gustavo Azzolin KTH, Royal Institute of Technology SE-164 40 Stockholm, Sweden gadcp@kth.se Abstract— It has for long been recognized that in multi-cell wide-band code division multiple access (WCDMA) networks the admission of a new session into the system can have undesirable impacts on the neighboring cells. Although admission control algorithms that take into account such multi-cell impacts have been studied in the past, little attention has been paid to multi-cell admission and rate control algorithms when traffic is elastic. In this work, we propose a model for multi-cell WCDMA networks to study the impact of admission and rate control algorithms on key performance measures such as the class-wise blocking and dropping probabilities, block error rates and the overall throughput. By means of simulation we compare the performance of multi-cell algorithms with that of a single cell algorithm. For voice traffic, we find that multi-cell algorithms improve the session drop and block error probabilities in both of the examined heterogenous (termed "Hotspot" and "Hotround") scenarios. When traffic is elastic, the multicell algorithms perform better in the examined homogeneous case. I. I NTRODUCTION The teletraffic behavior and the issue of admission control in code division multiple access (CDMA) networks have been the topic of research ever since CDMA started to gain popularity for military and commercial applications, see for instance Chapter 6 of [8]. The paper by Evans and Everitt used an M/G/∞ queue model to assess the uplink capacity of CDMA cellular networks and also presented a technique to calculate the outage probability [4]. It has been emphasized by several contributions that multi-cell admission control improves the resource utilization and reduces outages (session drops), see for instance [7]. These classical papers have focused on "rigid" traffic in the sense that elastic or best effort traffic whose bit rate can dynamically change was not part of the models. Subsequently, the seminal paper by Altman proposed a Shannon like capacity measure called the "best effort capacity" that explicitly takes into account the behavior of elastic sessions [1]. The importance of taking into account the relation between the number of admitted sessions and session drops (and their impacts on the Erlang capacity) in cellular networks in general and in CDMA in particular has been emphasized by several authors, see for instance [4] and more recently [3]. Session drops are primarily caused by outages, when the desired signal-to-noise ratio for a session stays under a predefined threshold during such a long time that the session gets interrupted. However, sessions can be dropped by a load control algorithm (typically located in the radio network controller in WCDMA) to preserve system stability. Session interruptions are perceived negatively by end users - more negatively than blocking a session - and therefore their probability should be minimized by suitable resource management (including admission control) techniques. The purpose of this paper is to develop a model that can be used to analyze the usefulness of multi-cell admission control (in terms of its impact on the blocking and dropping probabilities and also on the block error rate, BLER, as discussed in the paper) when traffic is elastic. The contribution of this work is the development of a model that can be used to analyze the impact of multi-cell admission control when some of the traffic classes tolerate a certain slowdown at the expense of increasing the holding time of the sessions. To the best of our knowledge, the modeling of elastic traffic and the analysis of the associated admission control algorithm in this type of dynamic multi-cell CDMA system is novel. We proceed along the following line. First, in Section II, we establish a basic relationship between the load factor (and thereby the noise rise) increase and the session-wise required peak and minimum bit rates when an elastic session is admitted into the system. Next, in Section III, we develop the so called stateful multicell model that explicitly takes into account the states (the number of in-progress sessions and their bit-rates) of all involved cells. Although the stateful model is useful, it is not practical in real systems due to computational aspects. Therefore, in Section IV we develop a feasible admission and rate control algorithm that can be used in real systems. This algorithm is analyzed by means of simulations in three seven- cell scenarios (termed homogeneous, hotspot and hotround) in Section V. We draw conclusions and summarize our findings in Section VI. II. MODELING ELASTIC TRAFFIC IN CDMA (THE SINGLE CELL CASE) A. Basic CDMA Equations Consider a single CDMA cell at which sessions belonging to one of K service classes arrive according to a Poisson arrival process of intensity λ k (k =1,...,K). Each class is characterized by a peak bit-rate requirement ˆ R k and an exponentially distributed nominal holding time with parameter μ k . When sending with the peak rate for a session, the required 1-4244-0353-7/07/$25.00 ©2007 IEEE This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2007 proceedings.