CONCLUSIONS

In this report, we have analyzed the relation among the physical layer bandwidth, the TCP layer throughput, delay and queue size when delivering the TCP traffic in the WCDMA system.

The developed analytical formulas can facilitate the design of queue sizes in the buffer of base station for different TCP throughput and delay requirements. Specifically, we propose an explicit rate change notification (ERCN) mechanism to dynamically change the queue size of the buffer in the base station based on the TCP/Physical cross-layer performance issues. Hence, when the radio link capacity in the physical layer is changed, the suitable queue size in the base station can be effectively adapted and the TCP sender can accordingly change the transmit data rates, thereby both the TCP delay and throughput requirements can be met in the varying radio channel conditions. In addition, we validate the accuracy of the analysis by simulation in the EURANE, the simulation results are similar to the results of analysis. Finally, we have numerical results to show the ERCN can improve the TCP throughput, delay and delay jitter when the WCDMA system adapts the wireless link data rate to the radio channel condition.

REFERENCES

[1] S. Shakkottai, T. S. Rappaport, and P. C. Karlsson, “Cross-layer design for wireless networks,” IEEE Communications Magazine, vol. 41, NO. 10, pp. 74 – 80, Oct 2003.

[2] F. Lefevre and G. Vivier, “Understanding tcp’s behavior over wireless links,” IEEE Vehicular Technology Conference.

[3] M. Sagfors, R. Ludwig, M. Meyer, and J. Peisa, “Buffer management for rate-varying 3g wireless links supporting tcp traffic,”

IEEE Vehicular Technology Conference, vol. 1, pp. 675 – 679, April 2004.

[4] ——, “Queue management for tcp traffic over 3g links,” IEEE Wireless Communications and Networking Conference, vol. 3, pp. 1663–1668, March 2003.

[5] H. Ekstrom and A. Schieder, “Buffer management for the interactive bearer in geran,” IEEE Vehicular Technology Conference, vol. 4, pp. 2505–2509, April 2003.

[6] S. Floyd, “Red wed page,” Aug 2001. [Online]. Available: http://www.aciri.org/floyd/red.html [7] R. Stevens, TCP/IP Illustrated. Addison-Wesley, 1994, vol. 1.

[8] M. Yavuz and F. Khafizov, “Tcp over wireless links with variable bandwidth,” IEEE Vehicular Technology Conference, vol. 3, pp. 1322 – 1327, Sept 2002.

[9] P. J. Ameigeiras, J. Wigard, and P. Mogensen, “Impact of tcp flow control on the radio resource management of wcdma networks,” IEEE Vehicular Technology Conference, vol. 2, pp. 977 – 981, May 2002.

[10] R. Cuny and A. Lakaniemi, “Voip in 3g networks: an end-to-end quality of service analysis,” IEEE Vehicular Technology Conference, vol. 2, pp. 930–934, April 2003.

[11] H. Holma and A. Toskala, WCDMA FOR UMTS. England: John Wiley and Sons, 2000.

[12] M. Kwon and S. Fahmy, “Tcp increase/decrease behavior with explicit congestion notification (ecn),” IEEE International Conference on Communications, vol. 4, pp. 2335 – 2340, May 2002.

[13] K. K. Ramakrishnan, S. Floyd, and D. Black, The Addition of Explicit Congestion Notification (ECN) to IP, ,RFC 3168, Sept 2001.

[14] J. Padhye, V. Firoiu, D. F. Towsley, and J. F. Kurose, “Modeling tcp reno performance: a simple model and its empirical validation,” IEEE/ACM Transactions on Networking, vol. 2, NO. 3, pp. 133 – 145, April 2000.

[15] “Enhanced umts radio acces network extension for ns-2,” Aug 2004. [Online]. Available: http://www.ti-wmc.nl/eurane/

[16] “The network simulator ns2,” Aug 2004. [Online]. Available: www.isi.edu/nsnam/ns/

Comparisons of Link Adaptation Based Scheduling Algorithms for the WCDMA System with High

Speed Downlink Packet Access

Li-Chun Wang and Ming-Chi Chen

I. INTRODUCTION

In order to satisfy the fast growing demand of the wireless packet data services, the concept of high speed downlink packet access (HSDPA) is proposed as an evolution for the wideband code division multiple access (WCDMA) system [1]. The goal of the WCDMA system with HSDPA is to support peak data rates from 120 kbps to 10 Mbps by adopting many advanced techniques, such as fast link adaptation, fast physical layer retransmission, and efficient scheduling techniques [2].

A fast link adaptation mechanism can enhance throughput performance by adapting modulation and coding schemes in the rapidly changing radio channel. The hybrid automatic repeat request (HARQ) technique can improve the radio link performance by combining retransmitted packets with previous erroneous packets. Scheduling is the key to achieving fairness in a shared channel for multiple users. Basically, a scheduling algorithm is to select a most suitable user to access the channel in order to optimize throughput, fairness, and delay performances.

Recently, scheduling has attracted much attention for wireless data networks because it can exploit the multi-user diversity [3] [4]. In the traditional voice-oriented cellular network, the fluctuation of the fast fading is viewed as a drawback. However, channel variations can be also beneficial to the wireless data network. Because data services can tolerate some delay, a scheduling mechanism can be designed to select the user with the highest channel peak and serve one user at a time in a time-multiplexing fashion. With a higher channel peak, a more efficient modulation/coding scheme can be applied to enhance data rates. In general, a larger dynamic range of channel variations can yield higher channel peaks, thereby delivering larger multi-user diversity gain.

In addition to throughput optimization, service delay and fairness are another two important aspects needed to be taken into account in designing a good scheduling algorithm. The reason why wireless scheduling can improve system throughput lies in the fact that data services can tolerate certain level of delay. Nevertheless, a constraint on the maximum service delay is still necessary. In wireless systems, mobile users are located at different locations with different channel conditions. If a scheduler always selects the user with the best channel condition, some users at the cell boundary, for instance, may never have a chance to access the system. Hence, how to design a scheduling algorithm to achieve high throughput subject to delay and fairness constraint becomes a crucial and challenging issue for the wireless data network.

In the literature, according to the considered channel models, wireless scheduling algorithms can be categorized into two major types. First, the wireless scheduling algorithms in [5] [6]

[7] considered a two-state on-off Markov channel model. Because of simplicity, the two-state Markov channel model is suitable to examine the fairness performance of scheduling algorithms.

However, using the simple two-state Markov channel has limitations in capturing actual radio channel characteristics. Second, some wireless scheduling algorithms like [1] [8] [9] [11] considered a more practical radio channel model with the emphasis on exploiting the multi-user diversity. In [1], the maximum carrier to interference ratio (C/I) scheduler is designed to assign the channel to the user with the best C/I. Obviously, the maximum C/I scheduler fully utilizes the multi-user diversity, but is an unfair scheduling policy. In contrast to the maximum C/I scheduler, a fair time scheduler (or called the round robin scheduling algorithm [8]) allocates the channel to users in sequence with equal service time. Clearly, the fair time scheduler does not consider the channel effect. In [2] [8]

[9] [10], the proportional fair scheduler was proposed for the IS-856 system and the WCDMA system. The proportional fair scheduler improves the fairness performance of the maximum C/I scheduler at the cost of lowering system throughput. However, in [11] it was pointed out that the proportional fair scheduling algorithm does not account for service delays. Thus, the authors in [11]

proposed the exponential rule scheduler to improve the delay performance of the proportional fair scheduler. It was proved that the exponential rule scheduler is throughput optimal in the sense of making a service queue stable [12]. In [13], it was concluded that the exponential rule scheduler is superior to the proportional fair scheduler since it provides excellent latency performance even with a slightly lower system throughput. Nevertheless, the factor of queue length is still not explicitly

considered in the exponential rule scheduling policy.

To our knowledge, a wireless scheduling policy with consideration of all the factors of channel variations, service delay, and queue length is still lacking in the literature. Consequently, we are motivated to develop such a wireless scheduling algorithm. The contributions of this work are two folds. The first contribution of this work is to propose a queue-based exponential rule scheduler to explicitly take account of all the factors consisting of queue length, service delay, and channel variations. We find that the queue-based exponential rule scheduler can further improve the fairness performance as compared to the original exponential rule scheduler, while maintaining the same throughput and delay performances.

Second, in the context of multi-type services, we suggest a fairness index to evaluate the fairness performance of the link adaptation based wireless scheduling algorithms, i.e., the maximum C/I, proportional fair, and exponential rule schedulers. This fairness index is modified from the one used in the two-state Markov channel based wireless scheduling algorithms [5] [6] [7] [14]. Interestingly, we find the fairness of current link adaptation based wireless scheduling algorithms [1] [8] [9] [11]

is not clearly specified. Thus, we are motivated to adopt a formal fairness index to evaluate the fairness performance of these link adaptation based wireless scheduling algorithms. Using this fairness index to compare scheduling algorithms is important, especially supporting multi-type services. For example, the fair time (or round robin) scheduler is viewed as the performance upper bound in terms of fairness for most link adaptation based wireless scheduling algorithms. However, this upper bound is only valid under the assumption that all users subscribe the same type of service any time. In the case of multi-type services, the fair time scheduler can only guarantee the equal access time for multiple users, but not ensure to satisfy the different requirements for different users.

Thus, even the fair time scheduler may not be the fairest scheduling algorithm in supporting the multi-type services. Thus, to support multi-type services, it is important to re-evaluate the fairness of these link adaptation based wireless scheduling algorithms based on a formal fairness index.

Our simulation results show that in the time-multiplexing fashion, the fairness performance of the exponential rule scheduler is very close to that of the fair time scheduler and both schemes are superior to the proportional fair scheduler. On the other hand, in the code-multiplexing fashion, we find that the exponential rule scheduler is only slightly better than the proportional fair rule, and both schemes are worse than the fair time scheduler.

TABLE I

MODULATION ANDCODINGSCHEMES IN THEHSDPA CONCEPT

Modulation and coding schemes (MCS) Modulation Effective code rate

MCS 1 QPSK 1/4

MCS 2 QPSK 1/2

MCS 3 QPSK 3/4

MCS 4 16-QAM 1/2

MCS 5 16-QAM 3/4

The rest of this part of report is organized as follows. Section II briefly introduces the background of the HSDPA concept in the WCDMA system. Section III describes existing link adaptation based scheduling algorithms. We discuss our new proposed queue-based exponential rule scheduling algorithm in Section IV. Simulation results are presented in Section V. Finally, we give our concluding remarks in Section VI.