Iccit 2007

Modeling and Performance of TCP in a MCCDMA System for 4G Communications Khan Md. Rezaul Hoque*, Riaz Raihan Haque†, Md. Akbar Hossain*, M.S.Feroz Farazi*,Gahangir Hossain** *

Dept. of ICT, University of Trento, Via Sommarive, 14 I-38100 POVO, Trento, Italy † Dept. of EEE, Islamic University of Technology, Gazipur, Bangladesh ** Dept. of CSE, Chittagong University of Engineering and Technology, Chittagong, Bangladesh Email: [email protected] Abstract—Market research finds that mobile commerce for 4G wireless systems will be dominated by basic human communication such as messaging, voice, and video communication. Because of its typically large bandwidth requirements, broadband communication is expected to emerge as the dominant type of traffic in 4G wireless systems. In this paper a new TCP based Multicarrier access technique named MC-CDMA for mitigating 4G requirements is proposed. This paper also presents analytical information regarding the transfer of TCP data flows on paths towards interconnected wireless systems, with emphasis on 4G cellular networks. The focus is on protocol modifications in face of problems arising from terminal mobility and wireless transmission. We advocate the use of TCP as the transport layer protocol for highspeed data in a Multi-Carrier CDMA (MC-CDMA) system for 4G wireless communications. Keywords— 4G, MC-CDMA, TCP, layer mechanisms.

I. INTRODUCTION The 4G will be a fully IP-based integrated system of systems and network of networks achieved after the convergence of wired and wireless networks as well as computer, consumer electronics, communication technology, and several other convergences that will be capable of providing 100 Mbit/s and 1 Gbit/s, respectively, in outdoor and indoor environments with end-to-end QoS and high security, offering any kind of services anytime, anywhere, at affordable cost and one billing [1]. The proliferation of the Transmission Control Protocol (TCP) in Internet communications today incites the research community to further extend its use in mobile and wireless networks. The ultimate goal is efficient and reliable TCP flows for Internet traffic over interconnected wired and wireless paths, where the wireless path suffers from additional problems due to higher BERs (Bit Error Rates) and frequent link changes. This primarily entails the treatment of protocol issues, but also additional interoperability in the network infrastructure. In the large-scale mobility case, cellular networks of the 4th generation (4G) are the most suitable candidates for support of Internet traffic, as they offer capacity for enhanced broadband data transfers, as well as improved transmission quality. In this paper, we describe the vision of the 4G systems focusing on major key technology such as MCCDMA system. The remainder of this paper is organized as follows. Section 2 and 3 provides a brief background on aspects of 4G cellular environment and MC-CDMA system pertinent to the subject of the paper. Section 4 provides related research in this field. Section 5 illustrates

problems of deploying TCP in 4G communication. Section 6 sets the tone for the following three sections by suggesting that performance improvements in TCP can be brought about by changes in any of the physical, link or transport layers. Section 6.1, 6.2 and 6.3 present modified physical layer, link layer and transport layer techniques respectively, which help in improving the performance of TCP in cellular wireless systems. Section 7 concludes the paper and highlights a few areas for future research. II. VIEW OF 4G MOBILE AND WIRELESS COMMUNICATION SYSTEMS 4G Mobile and wireless communication systems should support following functions: 1. Higher transmission rate up to 100Mbps 2. Flexible to advanced Internet, QoS control 3. Enhanced security 4. Seamless operation across networks 5. Multiple broadband access options in combined public and private networks including wireless LAN, wireless home link and ad-hoc network. 1G and 2G systems were voice communications, and digitized voice communications with some data communications, respectively, where a major difference was roaming between regions. 3G systems provide multimedia and wireless Internet at relatively high data rates, by utilizing packet switched services. However, significant paradigm shift should be taken into account for 4G systems, since wireless LAN, wireless MAN (WiMAX), wireless ad-hoc and sensor networks are becoming popular. Fig. 1 shows the evolution of networking and paradigm shift. Up to 2001, web-based service by using dial-up or always-on IP connection has

Fig.1 The evolution of networking and paradigm shift

been dominant. Now, mobile Internet is very popular and the driving force is mobile. The flexible and secure broadband seamless networking is the key to establish Ubiquitous network which is characterized by distributed computing, broadband and wireless, and peer-to peer for everything, and driving force is service. In our view, 4G systems are regarded as a “shopping mall type”, whereas 3G systems are “department store type”. Key issues for seamless operation are: 1. Service discovery and fast seamless connections/ services in the IP-based multi-modal access 2. Mobility management 3. IP multimedia services platform independent of radio access technology and underlying IP transport technology 4. Enhancement to support Human (H) to H, H to Machine (M) and M to M communications 5. Flexible introduction of new technologies into a system and service Fig. 2 shows view of the IP-based 4G mobile and wireless network architecture. IP-based backbone transport network supports multi-modal access among various wireless networks. As a lower middleware, the basic network management layer treats many functions related to multiple interface management, mobility management using mobile anchor point with buffer, security, QoS etc. As an upper layer of the services middleware, service support layer handles location, billing, media conversion, distribution etc. Application can be operated by using such

Fig. 2 View of the IP-based 4G mobile and wireless network architecture

Fig. 3 History of mobile communications systems in terms of adopted radio access technique

common service middleware. 4G systems are also characterized by the bandwidth to be allocated. In 2-5 GHz band, propagation loss is higher resulting in smaller cell size. Also, due to higher Doppler shift, more complex and robust synchronization and channel estimation techniques are needed. Key technologies being researched in physical layer are OFDM, multi-carrier CDMA (MCCDMA), multi-hop systems, MIMO and AAA, Time Division Duplex (TDD) CDMA, and downlink queuing and scheduling algorithm, routing protocol and distributed public key management for mobile ad-hoc networks in higher layers. III. MC-CDMA TECHNIQUES FOR 4G SYSTEMS Fig. 3 shows the evolution of mobile communications systems. In discussions about 2G systems in the 1980s, two candidates for the radio access technique existed, time division multiple access (TDMA) and CDMA schemes. Finally, the TDMA scheme was adopted as the standard. On the other hand, in the discussions about 3G systems in the 1990s, there were also two candidates, the CDMA scheme, which was adopted in the one generation older systems, and the OFDM-based multiple access scheme called band division multiple access (BDMA) [2]. CDMA was finally adopted as the standard. If history is repeated, namely, if the radio access technique that was once not adopted can become a standard in new generation systems, then the OFDM-based technique looks promising as a 4G standard. It is well known that the CDMA scheme is robust to frequency selective fading. Likewise, the OFDM scheme is a1so inherently robust to frequency selective fading. Thus the combination of OFDM and CDMA schemes may give batter performance. In 1993, Multi-carrier CDMA (MC-CDMA) system which is indeed a combination of OFDM and CDMA schemes was independently proposed by three different groups [4-6] almost simultaneously. Currently, an updated DS-CDMA technique is being used for the 4G mobile communication systems [7]. However the MC-CDMA looks more promising for the 4G standard. In DS-CDMA, Rake receivers are used, which have complexity in design and provide low performance. Thus, DS-CDMA is no longer the best choice available for the 4G standard. On the other hand, MC-CDMA system directly applies coding in the frequency domain to separate each user and thereby making it possible to support multiple users. This system is more effective in eliminating the frequency selective problem faced in the DS-CDMA system by not dividing the user bits into chip to uniquely encode them [7]. Instead, the user's codes are employed in different frequency and thus offer longer bit duration and make the signal to experience only flat fading. This is how it mitigates the problem of inter-chip interference [8]. Moreover, MC-CDMA system has lot of flexibility inherent in terms of system design that allows better spectrum utilization [9]. It also has the benefit of relatively simple receiver structure. IV. RELATED WORK IN THIS FIELD Numerous approaches have been proposed in the literature to optimize TCP performance in 4G wireless networks. These approaches can be broadly categorized as either TCP enhancement approaches or layer

optimization/modification approaches. The approach presented in this paper belongs to the latter category. Hence, notwithstanding their importance, we only briefly mention some of the works in this category. TCP enhancement consists of approaches that either introduce end-to-end TCP modifications or split the TCP connection with the help of an intelligent agent. A few examples of the former are TCP Westwood [10], TCP-Freeze [11] and the Eifel timer [12]. Examples of the latter are Snoop [6] and the ACK and Window-regulator [13,14]. We refer the reader to [15,16] for a more detailed survey. The intercede interference in frequency-selective Rayleigh fading channels in MC-CDMA is evaluated and the theoretical expressions for the desired-to-undesired signal power ratio (DUR) is derived by strictly considering the correlation property between subcarriers in [17]. The optimum spreading code assignment method is proposed in MCCDMA cellular systems over correlated frequencyselective Rayleigh fading channels based on the DUR imbalance among assigned spreading codes [18]. V. PROBLEMS OF DEPLOYING TCP IN 4G COMMUNICATIONS Typically, TCP is used in wired communication systems with very small errors probabilities. The error characteristics of wireless channels, however, differ significantly from that of wired channels. Therefore, TCP gives very poor performance if it is directly applied to a wireless communication system. Wired channel are characterized by miniscule packet loss probabilities and randomly spaced errors. In contrast, wireless channels are characterized by time varying packet loss probabilities that are generally much larger than for wired channels. Also, the errors are typically bursty on wireless channels [19]. Moreover, wireless channels are distinct and time varying between the Wireless Terminals (WTs), that is, the wireless link errors is location dependent. The variability of the wireless channel quality is due to the mobility of the WTs, fading effects, interference from other WTs, and shadowing. All of these effects degrade the channel performance significantly and have a significant impact on higher protocol layers. Numerous studies have found that TCP supports wireless Internet access only very inefficiently [20]. The key problems is that wireless channel errors lead to frequent expirations of the TCP retransmission timer, which are interpreted as congestion by TCP. VI. POSSIBLE SOLUTION OF TCP WITH MC-CDMA FOR 4G The performance metric of TCP in cellular environment is the average throughput, which is the same as in wired networks. However, the average throughput in the cellular case is not only dependent on the congestion in the network, but also on factors like bit error rate of the wireless medium, the cell handoff time and the cell resident time [21], [22]. Figure 4 below shows the protocol layers present at the various nodes in the network including the stationary host and the mobile host. The degradation in the TCP performance as mentioned earlier is mainly due to the lossy nature of the wireless link (between the mobile host and the base station) and also

Fig. 4 Protocol Stack Diagram

due to the intermittent disconnections caused by motion of the mobile host across cells. The physical layer between the BS and the mobile host can be improved to lower the packet loss rates and thus reduce the degradation of the TCP throughput. The link layer protocol that operates on the one-hop wireless link (between the BS and the mobile host) can also be modified to hide the wireless transmission losses using local recovery mechanisms (using link layer retransmission). A direct way to improve the performance of TCP will be to modify TCP itself, since it is the inherent assumptions of TCP that are the cause of its poor performance in the wireless environment [23]. We emphasize on modification of TCP in this paper depending upon the layer at which they are predominantly implemented, namely the physical, link or transport layer. We see that although IP layer is critical for providing inter-network mobility, it requires very little or no changes in directly effecting the performance of TCP. It will however be evident in some of the surveyed techniques that IP can aid the TCP layer in providing an awareness of host mobility, which proves useful in improving the performance. 6.1 Modified Physical Layer Techniques In this section, we present the Forward Error Correction and Interleaving techniques, which could be used at the physical layer in MC-CDMA systems. 6.1.1 Forward Error Correction (FEC) and Interleaving The physical layer in MC-CDMA uses convolutional and Turbo coding between the base station and mobile (in both directions) to achieve high coding gain. This helps in achieving a lower value of the packet error rate, this leads to a reduction in the burst errors, which is highly desirable. The FEC process is followed by the interleaving process, which provides time diversity as a further safeguard against burst errors. The basic function of the interleaver is to distribute one long burst of errors into many smaller bursts of correlated errors, spread across many different physical layer frames. This prevents the concentration of errors in any particular TCP window, which allows TCP fast retransmit mechanisms to offer better throughput, as shown in the simulations in [21]. MC-CDMA, owing to its wide-band signal, is able to discriminate between the various multipath arrivals [24]. The use of rake receivers (essentially a set of correlators) at both the base station and the mobile allows the combination of multipaths to form a strong signal. Thus, MC-CDMA uses multipath arrivals to its advantage in contrast to other wireless technologies, and hence suffers significantly lesser burst errors caused by multipath fading.

6.2 Modified Link Layer Techniques The 4G MC-CDMA systems [25] will be able to provide data rates of up to 2 Mbps per user, which is an appreciable improvement over the current day CDMA systems that can only provide up to 9.6kbps per user. An outcome of the increased data rates is that the transmission delays at BS are lowered which reduces the probability of interference between the TCP level and link level retransmissions significantly and thus makes the link layer mechanisms more viable. We feel that there is a strong need to further explore the link layer retransmission techniques in light of high data rates that will become available in the next generation (4G) MC-CDMA systems. 6.3 Modified Transport Layer Techniques 6.3.1 Fast Retransmission (Explicit Handoff Notification) Scheme Caceres et al. [26] present a seminal analysis of the impact of handoffs resulting from the host movement on the throughput and delay of the reliable transport layer protocols. In their experiment, they analyze the performance of Tahoe TCP in a 2 Mbps wireless LAN environment with the host motion simulated in software. This setup allows for the simulation of both the possible cellular scenarios i.e. overlapping cells and nonoverlapping cells with different beacon periods. The instants at which handoffs can be initiated are also precisely controlled using this host simulation software. The handoff process when the mobile host (MH) moves across non-overlapping cells is illustrated in the Figure 5. We will use base station (BS) in our description, which is the equivalent of Mobility Support Stations (MSS) mentioned in [26]. The BS defines each cell and also acts the gateway for routing the packets to and from the mobile hosts (MH) in the cell. The BS makes the MH aware of their presence by broadcasting periodic beacons over the

wireless medium. The BS for each cell runs Mobile IP protocol to provide support for inter-network mobility. The MH senses the new beacon as it crosses over to the new cell and sends a Greeting packet to the new base station (BS2). BS2 updates its system structures as it accepts the visiting MH and sends a Greeting Acknowledgment to the MH. BS2 also notifies the old base station (BS1) that the MH has moved and that it can be reached through BS2. BS1 adjusts its routing tables in order to forward to BS2 any packets that arrive for MH and acknowledges the handoff to BS2. Finally, BS2 acknowledges the completion of handoff (via Handoff Completion signal) to the MH. The MH, earlier on the receipt of Greeting Acknowledgment, had already made BS2 as its new gateway for routing the packets. Owing to the non-overlapping nature of the cells, discrete beacon period and at least two packet exchanges that are required before the MH can see BS2, there is a finite time window during which the data and acknowledgements cannot be reached to either of the two BS. This results in loss of almost the entire TCP transmission window for the transmitter at MH and the fixed host (FH). The loss is more significant for the data destined for MH due the longer routing inconsistencies, as BS1 does not know that MH has left the cell until it receives an explicit notification from BS2. Since the entire window is lost, the MH TCP sender can send no more data until the retransmission timeout occurs. This timeout throttles the congestion window and the retransmission timeout interval is also backed off exponentially. If the handoff is still in progress (which is likely in case the period between beacons is long, there may be multiple exponential back-offs which will have an effect of long pause in the active TCP connection where no data is being exchanged and the sender can do nothing but wait for the expiration of the retransmission timeout timer.

Fig. 5 Handoff Process

CONCLUSION We have proposed various techniques that provide mechanisms at different layers, which help in improving the performance of TCP in MC-CDMA cellular wireless networks. We feel that the physical and link layer techniques are absolutely necessary to deal with the harsh mobile radio environment. In addition, one or more transport layer techniques (may not be limited to the ones mentioned in the paper) may be combined judiciously with them to further improve the performance of TCP. The study of TCP performance in MC-CDMA environment has vast research potential, especially because the next generation cellular networks (4G) [25] have chosen MC-CDMA as the air interface protocol of choice. REFERENCES [1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

Young Kyun, Kim; Prasad, Ramjee. 4G Roadmap and Emerging Communication Technologies. Artech House, pp 12-13. ISBN 158053-931-9. Chuang, J., and N. Sollenberger, ‘‘Beyond 3G: Wideband Wireless Data Access Based on OFDM and Dynamic Packet Assignment,’’ IEEE Commun. Mag., Vol. 38, No. 7, pp. 78–87, July 2000. Y. W Cao, C. C. Ko. And T. T. Tjhung, “ A new multi-code multicarrier DS-CDMA system,” IEEE Global Telecommunication Conference, vol. 1 pp. 543-546, November 2001. Yee, N., J. P. Linnartz, and G. Fettweis, ‘‘Multicarrier CDMA in Indoor Wireless Radio Networks,’’ Proc. of IEEE PIMRC’93, Yokohama, Japan, pp. 109–113, September 1993. Fazel, K., and L. Papke, ‘‘On the Performance of ConvolutionallyCoded CDMA/ OFDM for Mobile Communication System,’’ Proc. of IEEE PIMRC’93, Yokohama, Japan, pp. 468–472, September 1993. Chouly, A., A. Brajal, and S. Jourdan, ‘‘Orthogonal Multicarrier Techniques Applied to Direct Sequence Spread Spectrum CDMA Systems,’’ Proc. of IEEE GLOBECOM’93, Houston, TX, pp. 1723–1728, November 1993. Rodger E. Ziemer and Roger L. Peterson, “Introduction to Digital Communication, Second Edition,” Prentice Hall, New Jersey, 2001. S. A. Zekavat and Carl R. Nassar, “Spectral sharing in multisystem environments via multi-carrier CDMA, ” Accepted in IEEE International Conference on Communications, ICC’03, Alaska, May 11-15, 2003. N. Nakagami, “The m-distribution, a general formula for intensity distribution of rapid fading,” In Statistical Method in Radio Wave Propagation, W. G. Hoffman, Ed. Oxford, U.K.:Pergamon, 1960.

[10] C. Casetti, M. Gerla, S. Mascolo, M. Y. Sanadidi, and R. Wang. TCP Westwood: end-to-end congestion control for wired/wireless networks. Wireless Networks, 8(5):467–479, 2002 [11] T. Goff, J. Moronski, D. S. Phatak, and V. Gupta. Freeze-TCP: A true end-to-end TCP enhancement mechanism for mobile environments. In Proc. IEEE INFOCOM, 2000. [12] R. Ludwig and R. H. Katz. The Eifel algorithm: Making TCP robust against spurious retransmissions. ACM SIGCOMM Comput. Commun. Rev., 30(1):30–36, 2000. [13] M. C. Chan and R. Ramjee. TCP/IP performance over 3G wireless links with rate and delay variation. In Proc. ACM MOBICOM, Atlanta, USA, Sep. 2002. [14] M. C. Chan and R. Ramjee. Improving TCP/IP performance over third generation wireless networks. In Proc. IEEE INFOCOM, Hong Kong, Mar. 2004. [15] C. Barakat, E. Altman, and W. Dabbous. On TCP performance in an heterogeneous network: A survey. IEEE Communications Magazine, pages 40–46, Jan. 2000. [16] H. Elaarag. Improving TCP performance over mobile networks. ACM Comput. Surv., 34(3):357–374, 2002. [17] T. Shono, T. Yamada, K. Kobayashi, K. Araki, I. Sasase, “Intercode interference and optimum spreading sequence in frequencyselective Rayleigh fading channels on uplink MC-CDMA”, Trans. of IEICE on Communications, Vol.E87-A, No.8, pp.1981-1993, Aug. 2004. [18] T. Shono, T. Yamada, K. Kobayashi, K. Araki, and I. Sasase, “Spreading code assignment for multicarrier CDMA system over frequency-selective fading channels”, Trans. On IEICE on Communications, Vol.E87-B, No.12, pp3734-3746, Dec. 2004. [19] G. Bao. Performance evaluation of TCP/RLP protocol stack over CDMA wireless link. Wireless Networks, pages 229-237, 1996. [20] M. Zorzi and R. R. Rao. The e_ect of correlated errors on the performance of TCP. IEEE Communications Letters, 1(September):127-129, 1997. [21] A.Chockalingam, M.Zorzi and R.R.Rao, “Performance of TCP on wireless fading links with memory,” Proc. IEEE, ICC’98, 595600, 1998. [22] A. Chan, Danny H.K. Tsang, and S. Gupta, "Impacts of Handoff on TCP Performance in Mobile Wireless Computing", Proceedings of IEEE International Conference on Personal Wireless Communications (ICPWC'97), India, December 1997. [23] G. Xylomenos and G. C. Polyzos. "Internet protocol performance over networks with wireless links". IEEE Network, 13(4):55-63, 1999. [24] William C.Y.Lee, “Overview of cellular CDMA,” IEEE Transactions on Vehicular Tehnology, Vol.40, No.2, May 1991. [25] 3GPP home page: http://www.3GPP.org. [26] R.Caceres and L.Iftode, “Improving the performance of reliable transport protocols in mobile computing environments,” IEEE Journal on Selected Areas in Communications, Vol. 13, No. 5, pp 850-857, June 1995