Traffic and Physical Layer Effects on Packet Scheduling Design in W-cdma Systems

In the framework of 3G systems, the adoption of efficient packet scheduling algorithms is needed to guarantee QoS as well as to provide high capacity. This letter reveals that for a proper system optimisation, different packet scheduling strategies should be used depending on the deployment scenario (macrocell, microcell, picocell), the link direction (uplink or downlink) as well as the service characteristics (continuous, bursty traffic). Introduction: W-CDMA packet based networks, such as the considered in the UTRA proposal [1], provide an inherent flexibility to handle the provision of future 3G mobile multimedia services. Radio Resource Management (RRM) entity is responsible for utilization of the air interface resources and, consequently, the adoption of efficient RRM algorithms is needed to guarantee QoS as well as to provide high capacity. Packet scheduling is one of the RRM functions that will help to achieve such objectives. In wireless environments, the problem of QoS provisioning for multimedia traffic has gained interest in the literature in recent years [2, 3]. However, little effort has been devoted to date in addressing the RRM topic to guarantee a given QoS in a packet driven environment such as the above mentioned in the framework of the 3G systems. It is worth mentioning that RRM functions can be implemented in many different ways, this having an expected impact on the overall system efficiency and on the operator infrastructure cost, so that definitively RRM strategies will play an important role in a mature UMTS scenario. Additionally, RRM strategies are not subject of standardisation, so that they can be a differentiation issue among equipment producers and operators. Packet scheduling strategies: The packet scheduling function shares the available air interface capacity between packet users. The packet scheduler can decide the allocated bit rates and the duration of the allocation. In W-CDMA this can be done in two ways [4], [5]: “code scheduling”, where a large number of users can have a low bit rate channel available simultaneously, and “time scheduling”, where capacity is given to a few number of users at each moment of time so that the user can have a very high bit rate but can use it only during small time periods. It is generally accepted that the design of an efficient packet scheduler is a difficult task that typically involves a large number of conflicting requirements (maximisation of throughput, minimisation of packet losses, etc.) which must be analysed and weighted before a balanced and fair solution can be found. The purpose of this letter is to point out that, in addition to the above issues, defining a packet scheduler needs to consider a number of system aspects as for example the service characteristics (in terms of traffic generation patterns). Also, the design of a layer 3 issue (Radio Resource) is influenced by a number of physical layer effects (layer 1). Consequently, a network operator pursuing to optimise the spectrum utilisation should implement different packet scheduling algorithms according to the specific scenario. System model and results: Characteristics of the cellular model are taken from [6]. The gaussian hypothesis and perfect power control are assumed for W-CDMA interference characterisation. Since the interest here is to focus on the scheduling process, traffic is supported through dedicated channels (DCH) [7]. In order to study the impact of the operation scenario, different orthogonality factor (ρ) values will be assumed [4] in order to represent different multipath propagation conditions (e.g. in microcells there is less multipath propagation and thus a better orthogonality). The orthogonality factor value also represents the link direction (for UTRA-FDD [1] ρ=1 for the uplink because codes used by different users are not orthogonal, while ρ≤1 for the downlink because signals are orthogonally transmitted from the base station by using OVSF codes). Note that the intracell interference is given by ρP, P being the received power from the own cell (e.g. ρ=0 means that W-CDMA signals are perfectly orthogonals and they do not cause interference to each other). The traffic model considered represents a typical interactive session, with several activity periods followed by thinking time periods. Within an activity period, several packet arrivals occur. In order to study the impact of the packet length on the packet scheduling strategies, the case of long packets is represented by an average length of 5000 bytes with a lognormal distribution, whereas the case of short packets is represented by an average 400 bytes [8]. Simulations are run for the case of long packets and short packets. In both cases several orthogonality factors are considered (ρ=1, 0.6 and 0.3) and results are obtained for “code scheduling” as well as for “time scheduling”. In order to summarise the results obtained, Table 1 and Table 2 focus only in terms of capacity (number of supported users under controlled performance) as a basis for comparisons. For long packets, a “time scheduling” policy offers better performance then a “code scheduling” policy denoting that it is not suitable to occupy the channel for a large period. However, if the orthogonality factor (which could be the case for the downlink or for micro/picocells) is reduced the “time scheduling” gain vanishes because this fact is favourable to the “code scheduling” capabilities. For short packets, a “code scheduling” policy clearly outperforms a “time scheduling” option, regardless of the considered scenario. Translating these results into service, link and environment parameters, it can be said that the traffic pattern (e.g. the service characteristics) is the dominant effect. For services generating relatively short packets the “code scheduling” approach is a suitable solution. For services generating relatively long packets the “time scheduling” approach is suitable for uplink direction (ρ=1), it would provide some gain in the downlink of a macrocell scenario (assimilated to ρ=0.6) and it would be similar to the “code scheduling” policy in the downlink of a micro/picocell (assimilated to ρ=0.3). Conclusions: The development of suitable packet scheduling mechanism have an impact on the overall system efficiency. This letter has revealed that for a proper system optimisation, different packet scheduling strategies should be used depending on the service characteristics (continuous or bursty traffic nature) as well as some system level issues (scenario and link layer) deriving from their associated physical layer characteristics (mainly in terms of degree of orthogonality). It is found that traffic characteristics is the most predominant effect. For short packets it is advantageous to use a “code scheduling” approach. For long packets, “time scheduling” strategy provides better performance, although the lower the multipath in the environment the lower the gain will be so that “code scheduling” would also be suitable for downlink in micro/picocells. Acknowledgements: This work is part of the ARROWS project, partially funded by the European Commission under the IST framework (IST 2000-25133) and by the Spanish Research Council under grant TIC2000-2813-CE. References: [1] http://www.3gpp.org [2] M. Naghshineh, A. S. Acampora, “Design and control of microcellular networks with QoS provisioning for data traffic”, Wireless Networks 3 (1997) , pp. 249-256 [3] I. F. Akyldiz, D. A. Levine, I. Joe, “A Slotted CDMA Protocol with BER Scheduling for Wireless Multimedia Networks”, IEEE/ACM Transactions On Networking, Vol. 7, No2, April 1999, pp. 146-158 [4] H. Holma (editor), WCDMA for UMTS, John Wiley & Sons, 2000 [5] O. Sallent, J. Pérez-Romero, et. al., “An Emulator Framework for a New Radio Resource Management for QoS Guaranteed Services in W-CDMA Systems”, IEEE Journal on Selected Areas in Communications, Vol. 19, No. 10, October 2001. [6] 3G TR 25.942 v.2.1.3, “RF System Scenarios” [7] 3GPP TS 25.211 v3.2.0, “Physical channels and mapping of transport channels onto physical channels (FDD)” [8] M. Bartoli et al., “Modelli di traffico E-mail e WWW per il servizio GPRS”, Internal Report Telecom Italia Lab (TILab), February 2000 Table captions: Table 1. Capacity for “code” and “time” scheduling in case of long packets. Table 2. Capacity for “code” and “time” scheduling in case of short packets.