INTERNATIONAL JOURNAL OF COMPUTERS AND COMMUNICATIONS

Issue 4, Volume 5, 2011

An adaptive call admission control scheme with load balancing for QoS enhancement in RoF based converged UMTS/WLAN
R. Shankar and P. Dananjayan

AbstractAs universal mobile telecommunications system (UMTS) and wireless local area network (WLAN) have complementary characteristics, convergence of UMTS and WLAN offer seamless wireless access service to users. In a converged UMTS/WLAN enhancing quality of service (QoS) for different services require call admission control and load balancing. This paper analyses the performance of wideband power based (WPB) and throughput based (TB) call admission control schemes in UMTS/WLAN. Further an adaptive call admission control scheme with load balancing for enhancing QoS in hybrid coupled and radio over fiber (RoF) based UMTS/WLAN convergence network is analysed. Simulation results show that adaptive call admission control with load balancing improves QoS in RoF based UMTS/WLAN compared to hybrid coupled UMTS/WLAN convergence network. KeywordsAdmission control, hybrid coupled, RoF, Converged UMTS WLAN, load balancing, QoS.

I. INTRODUCTION he 3 generation partnership project (3GPP) has been working on standardisation for integrating cellular and WLAN systems [1, 2], Network architectures for converging UMTS/WLAN systems can be grouped into two categories based on the independence between two networks as, tight coupling and loose coupling. In loose coupling architecture, the two networks are integrated beyond core network (CN) of UMTS and are connected through gateways of internet. Communication between networks are realised through standard internet protocol (IP) and mobility of mobile stations is managed through protocols such as Mobile IP. Loose coupling architecture enables the two networks deployed independently but results in longer delay for signaling and vertical handovers.

rd

Manuscript received August 30, 2011. R. Shankar is an Assistant Professor in the Department of ECE at Sri Manakula Vinayagar Engineering College, Pondicherry, India. (E-mail: [email protected]). P. Dananjayan is a Professor in Department of ECE, Pondicherry Engineering College, India. (E-mail: [email protected]).
271

In tight coupling architecture, the two networks are integrated at UMTSs CN, which has lower delay for signaling and vertical handover but has higher implementation complexity. Although above two schemes are used widely, they have their own disadvantages. To overcome their disadvantages hybrid coupling [3, 4] technique was proposed. In order to enhance

INTERNATIONAL JOURNAL OF COMPUTERS AND COMMUNICATIONS

Issue 4, Volume 5, 2011

QoS, a converging architecture using RoF technique is proposed [5]. Compared with the existing coupling schemes, RoF based convergence network can distribute the signaling and data load on both UMTS core network and core internet and have less handoff latency in vertical handoff process. When a mobile node (MN) is in UMTS cellular network as in Fig. 1, it directly connects with the base station (BS) by wireless link a. While the MN is in WLAN, it connects with remote AP by wireless link b. The remote AP transmits radio signal to WLAN server by RoF technique and builds up a connection with WLAN. RoF convergence architecture has following advantages: The BS controller of UMTS cellular network and server of WLAN are located nearer, as a result interchanging traffic between them will not burden core networks. MN is always connected with BS, the process of vertical handoff becomes simpler and communication is not disrupted during vertical handoff. The bandwidth of optical fiber is much broader than radio signal; consequently signals only occupy a little bandwidth of optical fiber. The hardware requirement for converging UMTS and WLAN networks is mainly to build dual mode user equipment (UE) which has capability to access either network. After such a dual mode UE (DMUE) [6] is available and software at each networks operational components are updated, a ubiquitous wireless environment with high data rate enabled in hot spot areas can be set up. To support various integrated services like voice, video, file transfer protocol (FTP), email etc. with QoS [7-9] in a UMTS/WLAN convergence network, resource provisioning like load balancing is a major issue in the converged environment. Call admission control with load balancing is an important solution because it is essential to maintain QoS in such converged network by having minimum call blocking and call dropping probabilities and at the same time efficiently utilise the available resources. The rest of the paper is organised as follows: admission control and load balancing is detailed in section 2, system model is described in section 3, simulation scenario and results are discussed in section 4, conclusion of this paper is presented in section 5. II. ADMISSION CONTROL AND LOAD BALANCING A. Wideband Power Based (WPB) Scheme Every time a new user seeks admission into the system, it adds certain amount of interference to the system. The criterion for uplink admission of the connection is based on comparison of interference the new user would add to system, if admitted, to an interference threshold value IT. This value should not be exceeded by the admission of a new user. If existing interference in system is IS and interference new user would bring to the system is I, then uplink criterion is (1) I S + I I T While uplink is interference limited, downlink in system is power limited. Considering downlink direction, user is
272

admitted if new total downlink transmission power does not exceed a predefined target value set by network operator: (2) PS + P > PT The load increase in downlink can be estimated on the base of initial power, which depends on distance from base station. The load increase depends on distance of mobile from base station. B. Throughput Based (TB) Scheme In uplink, a new user is admitted only if sum of existing uplink load factor [10] UL and increase in load factor UL does not exceed a predetermined threshold limit ULT. N (3) 1 UL = (1 + i ) w j =1 1+ Eb ( ) j . R j .v j N0 where N is number of stations, vj is active factor of station j at physical layer, Eb/N0 is signal energy per bit divided by noise spectral density, W is chip rate, Rj is bit rate of station j and i is other cell to own cell interference ratio seen by base station receiver. (4) UL + UL ULT The criterion in the downlink is similar to that of uplink, downlink load factor DL defined as N (E / N 0 ) j (5) DL = v j . b .[(1 j) + i j ] W Rj j =1 where j is orthogonality of channel of station j and ij is ratio of other cell to own cell base station received by user j. (6) DL + DL DLT where, increase in downlink load factor DL does not exceed a predetermined threshold limit DLT. The basis for this research was to apply WPB and TB schemes initially to UMTS and WLAN respectively in the converged UMTS/WLAN and changes them to TB and WPB in addition to load balancing, with varying percentages of data and voice users and study the performance. Sufficient preliminary results obtained on these schemes on the network consistently showed a trend of preferential treatment. It was observed that WPB scheme reduces voice blocking probability (pbvoice) more effectively than throughput based scheme whereas throughput based scheme works better in reducing data blocking probability (pbdata). Because, WPB is more power limited in downlink, voice users require lower power to be reached than data users. Hence the downlink forms a bottleneck for data users in WPB. For TB scheme uplink is capacity limited and data users are fewer in number at any point of time than voice users. Hence uplink forms a bottleneck for voice users in TB. Based on preliminary results, the need for a call admission control scheme that would minimise if not totally eliminates this bias was identified. An adaptive call admission control with load balancing that utilises both WPB and TB schemes depending on the number of users at a particular instant of time in the system is analysed for converged WLAN/ UMTS.

INTERNATIONAL JOURNAL OF COMPUTERS AND COMMUNICATIONS

Issue 4, Volume 5, 2011

C. Adaptive Call Admission Control (ACAC) Scheme In the ACAC scheme, updates of number of voice and data users in the system in periodic intervals of time , help determine which scheme WPB or TB it needs to switch to, by calculating number of each type of user present in system at the end of a previous period and number of each type of users estimated arrival in next period. If there are more voice users, ACAC switches to WPB and if there are more data users, it switches to TB. This prediction depends on , which is the parameter used to influence the number of predictions in upcoming period and , which influences total number of calls that have originated in the system since start-up. Simulations have shown that since video and FTP calls tend to persist in the system causing self-similarity, improves performance of ACAC. and vary between 0 and 1. The prediction for , and are clearly very critical. They can be found either adaptively or statistically. In this paper adaptive i.e., trial and error methods through many simulation runs are used. Though , and are determined heuristically, it has been observed that they are robust to variation in simulation scenarios for a certain link budget. The prediction is done in the following way for voice and data calls in the upcoming period: (7) V f = V n + (1 )V m + V total
D f = D n + (1 ) D m + D total

utilisation adopted is busyness ratio, which is defined as the ratio of time that the network is sensed busy. So stations are handovered to a WLAN network only when its busyness ratio Rb is less than a threshold, say Rth. Given the current busyness ratio Rb and its upper bound Rth, utility function for WLAN indicating available bandwidth to accommodate new stations is EP ( R th R b ) (12) LWLAN = TS which is derived from [12]. In equation (12) EP is average data packet size and Ts that depends on distributed coordinated function (DCF) inter frame space (DIFS) and short interframe space (SIFS) is the average time associated with a successful transmission. For the case that request to send/ clear to send (RTS/CTS) is not used, (13) T S = DIFS + T [ EP ] + SIFS + ACK and for the case that RTS/CTS mechanism is used, T S = DIFS + RTS + CTS + T [ EP ] + 2 SIFS + ACK (14)

where T[EP], RTS,CTS and ACK are time for transmitting a packet with payload size EP, transmitting RTS, CTS and ACK packet, respectively. III. SYSTEM MODEL DMUE is simulated in OPNET which can switch between UMTS and WLAN networks. The DMUE is different from dual mode terminal (DMT) [13] in which DMUE can be adopted in loose coupling interworking systems where UMTS and WLAN networks are connected by a router, whereas DMT is only applicable in tight coupling interworking systems. The protocols in UMTS and WLAN are relatively independent. The packets arriving at the router are routed according to subnet address of each network. Once packets are being delivered in UMTS or WLAN network, the delivery follows individual protocol. The main difference of UMTS and WLAN mobile station is in medium access control (MAC) layer [14] and physical layer. In DMUE, a new software layer, called IP switch layer is created, just below IP layer. Each DMUE has multiple IP addresses which access to different networks. When a data packet comes from the IP layer, the IP switch layer will make a decision based on WPB, TB or ACAC with load balancing and send out this data packet through corresponding physical layers. Each AP of WLAN and Node-B of UMTS uses TB and WPB initially to accommodate users present in converged area, which will provide maximum room for voice users and data users in UMTS and WLAN respectively. As time varies, when Node B of UMTS and AP of WLAN finds more data users and voice users, ACAC switches TB and WPB in WLAN and UMTS to WPB and TB respectively with load balancing function which will calculate available bandwidth in UMTS and WLAN in converged WLAN/UMTS for broadcasting. Once a station receives the load balancing function from either UMTS or WLAN network, it will compare received value with that of host network to decide
273

(8)

where Vf and Df are the predicted number of call arrivals in the coming period. Vn/Dn and Vm/Dm are originated and predicted voice/data calls in the previous period respectively and Vtotal/Dtotal is total number of voice/data calls that have originated in system since start up [11]. D. Load balancing function for the UMTS/ WLAN Given the current uplink load factor, UL, a load balancing function LUL(UL) can be constructed [6] (9) LUL ( UL ) = BUL .(1 UL ) T where BUL is maximum uplink bandwidth and T is total load factor Also, given the current downlink load factor, the linear load balancing function for downlink, (10) L DL ( DL ) = B DL .(1 DL ) T where BDL is the maximum downlink bandwidth. Thus, utility function of the uplink/downlink provides an estimation how much bandwidth is available in uplink/downlink. Then the load balancing function for UMTS network is designed as (11) LUMTS = LUL ( UL ) + L DL ( DL ) WLAN does not have explicit QoS control, achieves less throughput when the network is saturated than that when network is not saturated. As traffic load i.e., number of stations increases, severe collisions occur, which results in that the stations can barely transmit a packet successfully. Thus, WLAN network should be closely monitored such that the network is not overbusy. An indicator reflecting WLAN

INTERNATIONAL JOURNAL OF COMPUTERS AND COMMUNICATIONS

Issue 4, Volume 5, 2011

whether to swap network, after a random time, for each user before swap to avoid imbalance in network. Thus, resources of integrated networks are fully utilised meanwhile load balancing among networks is achieved. IV. SIMULATION SCENARIO AND RESULT Simulation of hybrid coupled and RoF based UMTS/WLAN convergence network is carried out using OPNET 14.5 [15], with the simulation parameters mentioned in table 1. Hybrid coupling is the combination of loose coupling and tight coupling and so WLAN router is connected with both IP cloud and UMTS CN, as in Fig. 2a. Here, AP router and WLAN router are connected with ethernet cable. This consists of Node-B, serving voice, video and FTP users.

WLAN data rate SIFS DIFS Phy header MAC header DATA packet RTS CTS, ACK Traffic frame size Utility gap Simulation time

54 Mbps 10 s 50 s 192 bits 224 bits 12000 bits + Phy header + MAC header 160 bits + Phy header 112 bits + Phy header 34560 bytes 10000 bits 1000 seconds

The RoF convergence network adopts RoF technique [5] that directly transmits radio signal on optical fiber. In this architecture, WLAN server and access point (AP) are connected with fiber distributed data interface (FDDI) cable as in Fig. 2b. The AP transmits the radio signal to WLAN server by RoF technique. Thus mobile node builds up a connection with WLAN. TABLE I SIMULATION PARAMETERS PARAMETERS BUL, BDL 1 Mbps
274

VALUES

INTERNATIONAL JOURNAL OF COMPUTERS AND COMMUNICATIONS

Issue 4, Volume 5, 2011

Both WLAN and UMTS networks are considered to be IPv6 based networks [16] and each node in the network have unique IP address. This consists of Node-B, serving voice, video and FTP users. The traffic generated is assumed to be similar as is the pattern of mobility. These systems have an approximately equal number of voice and data calls generated at the end of a period. Video and FTP users are referred as data users. Mobile nodes move with a velocity of approximately 40 km/hr. The input to system is based on traffic i.e., arrival rate or service rate that each type of service (ToS) provides rather than the average number of users. The output is observed as the blocking and the dropping probability. Voice users have ON and OFF periods Poisson distributed but with mean 0.65 and 0.35 seconds respectively. Voice and data users have durations that are uniformly distributed between 35 minutes and 1530 minutes respectively. FTP users have pareto distributed file size for self-similarity. Having a ToS helps to prioritize calls since priority can be considered to be another type of QoS. Voice, video and FTP users have ToS of interactive voice, global system for mobile (GSM) quality voice -priority 0, streaming multimedia, video conferencing light -priority 1 and best effort, file transfer light -priority 2 respectively.

DMUE has two IP addresses, one is for UMTS, and other is for WLAN. An application of UDP video uploading from each DMUE to server is applied. Each video frame size is 34560 bytes, and it has an interarrival time with a Poisson distribution.

Fig. 3a Throughput in hybrid coupled UMTS/WLAN

In case 1 of hybrid coupled and RoF based UMTS/WLAN convergence scenario, in the initiation, all the DMUEs
275

INTERNATIONAL JOURNAL OF COMPUTERS AND COMMUNICATIONS

Issue 4, Volume 5, 2011

randomly connect to either UMTS network or WLAN and switching never happen. In case 2, all the DMUEs initially connect to WLAN, the network access decision algorithm based on the load balancing with conventional call admission control (CAC) works in whole simulation duration. In case 3, all DMUEs in the converging area are admitted into UMTS or WLAN through ACAC with load balancing function to efficiently utilise the resource of converged WLAN/UMTS.

Fig. 4b Throughput in RoF based UMTS/WLAN

Fig. 3b Throughput in RoF based UMTS/WLAN

From the simulation result shown in Fig. 3a, 3b, 4a and 4b which are measured at the server side, overall throughput [1718] for 110Kb, 10Mb data in ACAC with load balancing approach is larger than one with random access and load balancing with conventional CAC in hybrid coupled and RoF based UMTS/WLAN convergence network respectively, further it can be inferred that the throughput is higher in RoF convergence network than hybrid coupled UMTS/WLAN.

Fig. 5a Blocking probability of hybrid coupled UMTS/WLAN

Fig. 4a Throughput in hybrid coupled UMTS/WLAN

Fig. 5b Blocking probability of RoF based UMTS/WLAN The results are obtained from an average of several runs of simulation for an uplink loading factor (UL) of 0.70 to 0.80 as most UMTS uses 0.75 as uplink and downlink load factors.

Fig. 5a and 5b shows the performance of TB, WPB, ACAC without load balancing and ACAC scheme with load balancing for blocking probability (PB) of hybrid coupled and RoF based UMTS/WLAN convergence network respectively.
276

INTERNATIONAL JOURNAL OF COMPUTERS AND COMMUNICATIONS

Issue 4, Volume 5, 2011

match that of the voice users. Due to this ACAC with load balancing is required, which will give tremendous improvement in performance and will work best under most conditions. Thus it is inferred that ACAC scheme with load balancing works best and enhances QoS [19-21] by minimising preferential treatment that is shown by WPB, TB schemes and ACAC without load balancing that lags in balancing load when used in hybrid coupled and RoF based WLAN/UMTS convergence network. V. CONCLUSION In this paper, wideband power based, throughput based call admission control scheme and an adaptive call admission control with load balancing is analysed to enhance QoS in hybrid coupled and RoF based WLAN/UMTS convergence network. The simulation results show that the adaptive call admission with load balancing performs significantly better in terms of QoS than throughput based, wideband power based, random network selection, adaptive call admission control without load balancing and load balancing scheme with conventional CAC. It is also observed that when using adaptive call admission control with load balancing, RoF based UMTS/WLAN convergence network outperforms hybrid coupled convergence network in terms of QoS. REFERENCES
3GPP TK 22.934, Feasibility study on 3GPP system to wireless local area network (WLAN) interworking, v. I .0.0, Release 6, Feb. 2002. [2] 3GPP, Group Services and System Aspects; 3GPP Systems to Wireless Local Area Network Interworking, System Description (Release 6), TS 23.234, v. 6.2.0, Sept. 2004. [3] R. Shankar and P. Dananjayan, An Adaptive Call Admission Control Scheme with Load Balancing for QoS Enhancement in Converged UMTS/WLAN, Proceedings of 5th International Conference on Communications and Information Technology, Corfu Island, Greece, July 14-16, 2011 [4] R. Shankar, Timothy Rajkumar K and P. Dananjayan, Security enhancement with optimal QoS using EAP-AKA in hybrid coupled 3GWLAN convergence network, International Journal of UbiComp (IJU), Vol.1. No. 3, pp. 31-42, July 2010. [5] Mingxin Chen, Gan Liu, Guangxi Zhu and Xu Zhu, A novel internetworking architecture between WLAN and UMTS cellular networks using RoF technique, Proceedings of IEEE International Conference on Circuits and Systems for Communications, Shangai, China , pp. 693-697, May 2008. [6] Yu Zhou, Yanxia Rong, Hyeong-Ah Choi, Jae-Hoon Kim, JungKyu Sohn, and Hyeong In Choi, Utility-Based Load balancing in WLAN/UMTS Internetworking Systems, IEEE Radio and Wireless Symposium, Orlando, FL , pp. 587 590, Jan. 2008. [7] Abdullah Alwehaibi, Anjali Agarwal, Michel Kadoch, and Ahmed Elhakeem, Accommodations of QoS DiffServ Over IP and MPLS Networks, WSEAS Trans. on Systems, pp.252-257, vol.2, issue 2, April 2003. [8] Capone and F. Martignon, Quality of Service Routing of Bandwidth Guaranteed Connections in MPLS Networks, Proceedings of 7th WSEAS International Conference on Communications, Corfu, Greece, July 2003. [9] O. Badarneh, M. Kadoch and A. Elhakeem, QoS Multilayered Multicast Routing Protocol for Video Transmission in Heterogeneous Wireless Ad Hoc Networks, WSEAS Trans. on Comp., Issue 6, Vol.7, pp. 680-693, June 2008. [10] H. Holma and A. Toskala, WCDMA for UMTS Radio Access for Third Generation Mobile Communications, John Wiley & Sons, 3rd edition, 2004
277

Fig. 6a Dropping probability of hybrid coupled UMTS/WLAN

[1]

Fig. 6b Dropping probability of RoF based UMTS/WLAN

Fig. 6a and 6b show the performance of TB, WPB, ACAC without load balancing and ACAC scheme with load balancing for dropping probability (PD) of hybrid coupled and RoF based UMTS/WLAN convergence network respectively. It is observed that ACAC with load balancing has the lowest values and they remain stable through different UL values. This is attributed to the fact that call admission control is a bidirectional scheme. Increase in UL might cause an increase in admitted users but might not meet requirement for downlink and viceversa. Also, PB and PD values of ACAC with load balancing are statistically different from WPB, TB and ACAC in hybrid and RoF based UMTS/WLAN convergence network. It is observed that RoF based UMTS/WLAN convergence network has linear variation in the blocking and dropping probability than hybrid coupled UMTS/WLAN in addition to reduced Pb and Pd that enhances QoS of the system. In a scenario where there are a dominant number of voice users or data users, ACAC will work as good as WPB or TB respectively but lags in balancing the load between WLAN/UMTS. But it is not seen in todays trend that UMTS systems where traffic generated by data users are rising up to

INTERNATIONAL JOURNAL OF COMPUTERS AND COMMUNICATIONS

Issue 4, Volume 5, 2011

[11] Kamala Subramaniam and Arne A. Nilsson, An Analytical Model for Adaptive Call Admission Control Scheme in a heterogeneous UMTSWCDMA System, IEEE International Conference on Communications ICC, Seoul, Korea, pp.3334 3338,Vol. 5, May 2005. [12] H. Zhai, X. Chen, and Y. Fang, How Well Can the IEEE 802.11 Wireless LAN Support Quality of Service, IEEE Trans. Wireless Communication, vol. 4, no. 6, pp. 3084 - 3094 Nov. 2005. [13] F. Siddiqui, S. Zeadally and S. Fowler, A Novel Architecture for Roaming between 3G and Wireless LANs, 1st Intl Conf. on Multimedia Services Access Networks, Orlando, FL, US A, pp.101-105, June 2005. [14] Tomislav Shuminoski and Toni Janevski, Novel Adaptive QoS Provisioning in Heterogeneous Wireless Environment, International Journal of Communication Networks and Information Security (IJCNIS), Vol. 3, No. 1, pp.1-7, April 2011. [15] OPNET network simulator, http://www.opnet.com/ [16] C. Liu and C. Zhou, HCRAS: A novel hybrid internetworking architecture between WLAN and UMTS cellular networks, Proceedings of IEEE Consumer Communications and Networking Conference, Nevada, USA pp. 374-379, January 2005. [17] J Zhu and Abraham O. Fapojuwo. A New Call Admission Control Method for Providing Desired Throughput and Delay Performance in IEEE802.11e Wireless LANs. IEEE Transactions on Wireless Communications, Vol. 6, No. 2, pp. 701-709, February 2007. [18] Y. Zhou, Y. Rong, H.-A. Choi, J.-H. Kim, J. K. Sohn, and H. I. Choi,"A Dual-Mode Mobile Station Modules for WLAN/UMTS Internetworking Systems," Proc. OPNETWORK 2007, Washington, DC, August 2007. [19] G. Sivaradje and P.Dananjayan "A Flexible Channel Allocation with Resource Prediction for Mobile Multimedia Networks" WSEAS Transactions on Computers, Issue 5, Volume 3, pp 1418-1424, November 2004.

[21] Ouldooz Baghban Karimi, Mahmood Fathy, "Adaptive end-to-end QoS for multimedia over heterogeneous networks", Computers & Electrical Engineering vol.31 no.1 pp.45-55, 2010. [20] Wei Song, Hai Jiang and Weihua Zhuang Performance analysis of the WLAN-First scheme in Cellular/WLAN interworking, IEEE transactions on wireless communications, Vol.6, Issue 5, pp. 19321952, May, 2007 R. Shankar received Bachelor of Engineering in 2001 and Master of Technology in 2006 in Electronics and Communication Engineering (ECE) from Bharathidasan University, Trichy and Pondicherry Engineering College, Pondicherry, India respectively. He is pursuing his Ph.D. in the Department of ECE, Pondicherry University. He is currently working as Assistant Professor in the Department of ECE at Sri Manakula Vinayagar Engineering College Pondicherry, India. His research interests include wireless communication, computer networks and convergence network. P. Dananjayan received Bachelor of Science from University of Madras in 1979, Bachelor of Technology in 1982 and Master of Engineering in 1984 from Madras Institute of Technology, Chennai and Ph.D. degree from Anna University, Chennai in 1998. He is working as a Professor and Chairman in the department of ECE and CBCS respectively in Pondicherry Engineering College, India. He also has been as visiting Professor to Asian Institute of Technology, Thailand. He has more than 100 publications in National and International Journals. He has presented more than 150 papers in National and International conferences. He has guided 11 Ph.D candidates and is currently guiding 6 Ph.D students. His areas of interest include spread spectrum techniques and wireless communication, wireless adhoc and sensor networks.

278