Dynamic Time-Threshold Based Scheme for Voice Calls in Cellular Networks

Transcription

1 Dynamic Time-Threshold Based Scheme for Voice Calls in Cellular Networks Idil Candan and Muhammed Salamah Computer Engineering Department, Eastern Mediterranean University, Gazimagosa, TRNC, Mersin 10 Turkey Abstract. This paper presents a dynamic bandwidth allocation scheme for new and handoff voice calls in cellular networks. The main idea of the new scheme is based on monitoring the elapsed real time of handoff calls and according to both a time threshold (t e ) and a dynamically changing bandwidth threshold (B t ) parameters, a handoff call is either prioritized or treated as a new call. Also in this paper, we introduce a crucial general performance metric Z that can be used to measure the performance of different bandwidth allocation schemes and compare them. Z, which is a performance/cost ratio, is a function of the new call blocking probability, handoff call dropping probability and system utilization all together. The results indicate that our scheme outperforms other traditional schemes in terms of performance/cost ratio, and maintains its superiority under different network circumstances. 1 Introduction With the increasing popularity of wireless communication systems, a satisfactory level of quality of service (QoS) should be guaranteed in order to manage the incoming calls more efficiently. However, establishment and management of connections are crucial issues in QoS-sensitive cellular networks due to user mobility. A wireless cellular network consists of a large number of cells, mobile users with various movement patterns and a number of applications. A cell involves a base station (BS) and a number of mobile stations (MSs). Handoff is the mechanism that transfers an ongoing call from one cell to another as a user moves through the coverage area of a cellular system. In literature, several approaches, like in [1-6] have been proposed to give priorities to handoff calls. In [1], resource reservation is made according to the staying time and switching time of an MS in order to increase bandwidth utilization. Static guard channel scheme (SGCS) in [2] gives priority to handoff calls by exclusively reserving fixed number of channels for handoff calls. Although SGCS decreases handoff dropping probability (P d ), it increases new call blocking probability (P b ) and may not utilize the system efficiently. The dynamic guard channel scheme (DGCS) in [2] is a variant of SGCS scheme where the number of guard channels is changed dynamically in order to improve the performance of the system. On the other hand, Y. Koucheryavy, J. Harju, and V.B. Iversen (Eds.): NEW2AN 2006, LNCS 4003, pp , Springer-Verlag Berlin Heidelberg 2006

2 190 I. Candan and M. Salamah according to the fully shared scheme (FSS), discussed in [2], all available channels in the cell are shared by handoff and new calls. Thus, FSS scheme minimizes the new call blocking probability and maximizes system utilization. However, it is difficult to guarantee the required dropping probability of handoff calls. Usually, SGCS and DGCS schemes are preferred by users since they decrease P d, and FSS scheme is preferred by service providers since it maximizes system utilization. As mentioned before, most of the schemes like in [3-15] prioritize handoff calls at the expense of blocking new calls. Their claim is forced termination of ongoing calls is more annoying than blocking of newly originating calls. We believe that this is true to some extent, as the annoyance is a fuzzy term which depends on the elapsed time of the ongoing call. For example, dropping an ongoing voice call is very annoying if it does not last for a moderate duration, whereas it is not that much annoying if it is approaching to its end. Motivated with these arguments, we introduce a novel dynamic bandwidth allocation scheme for voice calls which is based on fairness among calls and outperforms the SGCS, DGCS and FSS schemes. The main idea of our Dynamic- Time-Threshold-based Scheme (DTTS) is based on monitoring the elapsed real time of voice handoff calls and according to both a time threshold (t e ) and a dynamically changing bandwidth threshold (B t ) parameters, a handoff call is either prioritized or treated as a new call. 2 Dynamic Time-Threshold-Based Scheme (DTTS) In our scheme, we focus on a single cell as a reference cell in a cellular wireless network. We assume that the arrival traffic at the BS is of voice type, the cell has a total capacity B of 100 bandwidth units (BUs), and each call requires one unit. Fig. 1 below shows the bandwidth allocation of the DTTS scheme for new and handoff voice calls. The bandwidth threshold (B t ) dynamically changes between a fixed bandwidth threshold (B tfixed ) and the total capacity of the system (B) according to the time threshold t e values (i.e. B tfixed < B t < B). A new voice call or a nonprioritized handoff voice call is served if the amount of occupied bandwidth is less than B t BUs upon its arrival. However, a prioritized handoff voice call is served as long as the amount of occupied bandwidth is less than the total capacity which is B BUs. 0 B tfixed B t B New Voice and Non-Prioritized Handoff Voice Calls Prioritized Handoff Voice Calls Fig. 1. Bandwidth Allocation of the DTTS Scheme

3 Dynamic Time-Threshold Based Scheme for Voice Calls in Cellular Networks 191 Fig.2 below shows the flowchart of processing a voice call in the DTTS scheme. According to DTTS, handoff calls that have elapsed real time smaller than time threshold t e are prioritized. That is, such handoff calls are accepted as long as the amount of occupied bandwidth in the cell is smaller than the total capacity, B. On the other hand, handoff calls that have elapsed real time greater than or equal to time threshold t e, and new calls are treated according to the guard policy. That is, such calls are accepted as long as the amount of occupied bandwidth in the cell is less than a bandwidth threshold B t, where B t dynamically changes depending on the t e values. If the time threshold t e is large, the number of prioritized calls increases and the number of non-prioritized calls decreases. Therefore, the bandwidth threshold B t may be decreased. On the contrary, if the time threshold t e is small, the number of prioritized calls decreases and the number of non-prioritized calls increases. Therefore, the bandwidth threshold B t may be increased in order to increase the performance of the system. Since B t is directly proportional to t e, it can be defined as Bt B fixed Bt = te + B 2µ n where B tfixed is taken as 90 units [16] and µ n is the average service rate of a call. B t is dynamically changing between B tfixed and B units. (1) Voice call Handoff call? New call Non-prioritized call Elapsed time of the call t e? Yes Yes Occupied bandwidth < B t? Prioritized call No Call acceptance & bandwidth allocation No Occupied bandwidth < B? Yes Call blocking or dropping No Call dropping Fig. 2. A Voice Call Processing Flow Diagram In our previous work [17], we proposed a static time-threshold-based scheme (STTS) for voice calls where handoff voice calls are prioritized or treated as new calls according to a time threshold parameter t e. In the STTS scheme, prioritized handoff

4 192 I. Candan and M. Salamah calls are accepted as long as the amount of occupied bandwidth is smaller than total capacity, B. However, new calls and non-prioritized handoff calls are accepted as long as the amount of occupied bandwidth is smaller than a fixed bandwidth threshold B t. In the STTS scheme, the bandwidth threshold B t is not changed dynamically depending on the time threshold t e. That is, bandwidth threshold B t is fixed (B t = 90 units) in the STTS scheme. However, in the DTTS scheme presented in this paper, B t is dynamically varying between B tfixed and B bandwidth units as illustrated in both Fig. 1 and equation (1). 3 Simulation Parameters and Performance Metrics The simulation has been performed using the Ptolemy simulation tool, developed by the University of California at Berkley [18]. During simulation, more than 30 runs are taken for each point in order to reach 95% confidence level. As we mentioned before, we concentrate on a single reference cell. The interarrival times of new voice, handoff voice calls are assumed to follow Poisson processes with means 1/λ n and1/λ h respectively. The call holding times also follow exponential distribution with means 1/µ n for new calls, and 1/µ h for handoff calls. The average service time is defined as 1 µ λ 1 λ h n = + (2) λh + λn µ h λh + λn µ n The normalized offered load of the system (in Erlang) is defined as ρ Bµ 1 λ + n λh = (3) The mobility (γ ) of calls is a measure of terminal mobility and is defined as the ratio of handoff call arrival rate to new call arrival rate, and can be written as λ h γ = (4) λn The simulation input parameters used are given in following table (Table 1). The design goals of handoff schemes should also include minimizing the Grade of Service (GoS) cost function. Although sophisticated cost functions have been proposed [19], in practice, a simple weighted average is useful for most design purposes. The weighted sum of the new call blocking probability, prioritized handoff dropping probability (P d1 ) and non-prioritized handoff dropping probability (P d2 ) is introduced as a measure of grade of service (GoS) and can be defined as GoS P + k P + P = (5) b d1 d 2

5 Dynamic Time-Threshold Based Scheme for Voice Calls in Cellular Networks 193 Table 1. Simulation Parameters Mobility(γ ) 3 Load ( ρ ) Time threshold values (t e ) Total bandwidth (B) Bandwidth threshold (B t ) Fixed bandwidth threshold (B tfixed ) Arrival rate of handoff calls λ ) ( h Arrival rate of new calls ( λ ) New call average service time (1/µ n ) Handoff call average service time (1/µ h ) Average elapsed time of a handoff call n 0.9 Erlangs 30, 60, 90, 120, 150 and 180sec. 100 units units 90 units 0.6 calls/sec 0.2 calls/sec 180 sec. (Exp. Dist.) 180 sec. (Exp. Dist.) 90 sec. (Unif. Dist.) where k is the penalty factor used to reflect the effect of the handoff dropping over the new call blocking in the GoS cost function. A penalty of 5 to 20 times is commonly recommended [20]. In accordance with our proposed scheme, we used the penalty for the prioritized handoff calls, whereas the non-prioritized handoff calls have the same weight as the new calls. Of course, from the mobile user s point of view, the objective is to minimize the GoS cost function in order to improve the performance of the system. Therefore, the performance of a system can be defined as Performance 1 GoS = (6) Another objective, from the service provider s perspective is to decrease the cost by increasing utilization of the system. Therefore, the cost of a system can be defined as Cost 1 Utilization = (7) In order to make a fair balance between both user satisfaction and service provider satisfaction, a crucial performance metric Z is introduced to measure the performance of different bandwidth allocation schemes and compare them. Z can be defined as

6 194 I. Candan and M. Salamah Performance Z = (8) Cost It is clear that, Z is a function of new call blocking probability, handoff dropping probability and system utilization all together. Of course, the design goals of a handoff scheme are increasing the performance and decreasing the cost, which means maximizing Z. 4 Performance Results The proposed scheme is evaluated for different bandwidth threshold (B t ) values. The performance measures obtained through the simulation are the blocking probability of new voice calls (P b ), the total dropping probability of handoff voice calls (P d ) which includes both prioritized and non-prioritized handoff voice call dropping probabilities, grade of service (GoS) and Z. Fig. 3 below shows the relationship between B t and t e values that are calculated according to equation (1). It is clearly seen that B t decreases as t e increases. This is because, as t e increases, the number of prioritized handoff calls increases and therefore the guard bandwidth units should be enlarged to reserve enough bandwidth for these calls which can be achieved by decreasing B t. Bandw idth threshold,bt Time threshold, te (sec.) Fig. 3. Bandwidth threshold (B t ) versus time threshold t e Fig. 4 below shows the new call blocking probability (P b ) versus bandwidth threshold B t for different time threshold (t e ) values. It is seen that as B t increases, P b decreases for all t e values. Since the number of guard bandwidth units reserved for prioritized handoff calls decreases as B t increases, and therefore the chance of accepting a new call increases. The highest and lowest P b is obtained for t e =90 sec and t e =30 sec. The reason is that, when t e =90 sec. the number of prioritized handoff calls is high compared to the number of prioritized handoff calls when t e =30 sec. That is, when t e =90 sec, the number of guard bandwidth units reserved for prioritized handoff calls is higher than those reserved when t e =30 sec. It is worth to mention that there is no serious degradation in terms of P b for all t e values.

7 Dynamic Time-Threshold Based Scheme for Voice Calls in Cellular Networks New call blocking probability (Pb) Pb(DTTS, te=90) Pb(DTTS, te=60) Pb(DTTS, te=30) Fig. 4. New call blocking probability (P b ) versus bandwidth threshold (B t ) for different time threshold (t e ) values Fig. 5 below shows the handoff call blocking probability (P d ) versus bandwidth threshold B t for different time threshold (t e ) values. It is seen that as B t increases, P d decreases for all t e values. This result is interesting since we know that static or dynamic guard channel schemes(sgcs and DGCS) improves P d at the expense of increasing P b and degrading system utilization. On the other hand, fully shared scheme (FSS) improves P b and system utilization at the expense of increasing P d. However, our dynamic time-threshold scheme shows improvements in terms of both handoff dropping (P d ) and new call blocking (P b ) probabilities. From Fig.4 and Fig.5, it is also seen that the DTTS scheme shows good performance in terms of P b and P d for t e 90 sec. Therefore, it is recommended to set t e to a value less than or equal to the average service time of a handoff call, (½)µ h. Handoff call dropping probability (Pd) Pd(DTTS, te=60) Pd(DTTS,te=90) Pd(DTTS, te=30) Fig. 5. Handoff call dropping probability (P d ) versus bandwidth threshold (B t ) for different time threshold (t e ) values Fig. 6 below shows the grade of service (GoS) cost function versus bandwidth threshold B t for DGCS, DTTS and FSS schemes for k=10. The FSS gives the worst performance and it is almost constant as expected. It is very clear that our DTTS scheme gives the best GoS compared with the others. The DTTS scheme shows improvement over the DGCS and FSS schemes for all B t values. For example, DTTS improvements over DGCS and FSS schemes reach 135% and 230% at B t = 97 units respectively. It is worth to mention that similar trends of GoS as in Fig. 5 are observed for other values of k and better results are obtained for t e =30 and t e =60 sec.

8 196 I. Candan and M. Salamah 3.50E-01 Grade of service (GoS) 3.00E E E E E E E+00 GoS(DGCS,k=10) GoS(DTTS,te=90,k=10) GoS(FSS) Fig. 6. Grade of service (GoS) versus bandwidth threshold (B t ) for different schemes 1.40E+03 Perform ance/cost ratio (Z) 1.20E E E E E E E+00 Z(DTTS,te=30) Z(DTTS,te=90) Z(DTTS,te=60) Fig. 7. Z versus bandwidth threshold (B t ) for dynamic TTS scheme (DTTS) Perform ance/cost ratio (Z) Z(STTS,te=30) Z(STTS,te=90) Z(STTS,te=60) Z(DTTS,te=30) Z(DTTS,te=90) Z(DTTS,te=60) Fig. 8. Z versus bandwidth threshold (B t ) for dynamic TTS scheme (DTTS) and fixed TTS scheme (STTS) Fig. 7 below shows the performance/cost ratio (Z) versus bandwidth threshold B t for DTTS scheme under different t e values. It is clearly observed that, Z has the highest value when t e =30 sec. It is worth to mention that Z does not show a serious deterioration for other t e values. Fig. 8 below shows the performance/cost ratio (Z) versus bandwidth threshold B t for static time-threshold (STTS) and dynamic time-threshold (DTTS) schemes for

9 Dynamic Time-Threshold Based Scheme for Voice Calls in Cellular Networks 197 different t e values. It is clear that the Z value of DTTS scheme is higher than Z value of STTS scheme for all t e values. Fig. 9 shows the performance/cost ratio (Z) versus bandwidth threshold B t for SGCS, DGCS, FSS and DTTS schemes. It is clear that, our DTTS scheme outperforms the other schemes, since it has the highest Z for all B t values. As expected the Z value of the FSS is constant with respect to B t, because all channels are shared between new and handoff calls. The Z value of the SGCS is also constant since it has a fixed B t value which is 90 units. The DTTS scheme maintains its superiority over SGCS, DGCS and FSS scheme for all B t values. For example, DTTS improvements over SGCS, DGCS and FSS schemes reaches 200%, 134% and 228% at B t =97 units respectively. It is worth to mention that the DTTS scheme maintains its superiority for other values of t e. Performance/cost ratio (Z) 1.20E E E E E E E+00 Z(SGCS) Z(DGCS) Z(FSS) Z(DTTS,te=90) Fig. 9. Z versus bandwidth threshold (B t ) for dynamic TTS (DTTS), dynamic GCS (DGCS), static GCS scheme (SGCS) and fully shared (FSS) schemes 5 Conclusion In this paper, we have proposed and analyzed the performance of a new dynamic time-threshold based bandwidth allocation scheme (DTTS) for voice calls in cellular networks. The proposed DTTS scheme relies on fairness among new calls and handoff calls that may tolerate uncritical dropping. It is well known that the SGCS and DGCS aim to increase user satisfaction by decreasing the dropping probability; and the FSS aims to increase service provider satisfaction by decreasing the blocking probability and increasing bandwidth utilization. It is concluded that the proposed scheme outperforms these two extreme schemes. Hence, both users and service providers will be satisfied in the DTTS scheme. We also introduced a general performance metric Z as a performance/cost ratio that can be used to compare different bandwidth allocation schemes. The results show that the DTTS scheme maintains its superiority over STTS, SGCS, DGCS and FSS scheme for all bandwidth threshold values. It is worthy to mention that DTTS can be implemented easily since it uses the elapsed real time of a call which is already recorded in all systems for billing and other purposes. There are a number of issues that will be addressed in our future research. We are currently working on the analytical modeling of the proposed DTTS scheme. What is more, since the proposed scheme is applicable for FDMA systems, we are going to

10 198 I. Candan and M. Salamah improve it in order to be applicable for CDMA systems. In addition, we are going to apply a reverse strategy of the proposed DTTS scheme for data handoff calls, such that the data handoff calls that have elapsed time greater than time threshold t e are going to be prioritized. This is because; dropping an ongoing data call is very annoying if it is approaching to its end, whereas it is not that much annoying if it has just started. We believe that our scheme can be considered as a tool for service-balance in multimedia wireless networks. References 1. G. S. Kuo and P.C. Ko: A Probabilistic Resource Estimation and Semi-Reservation Scheme for Flow-Oriented Multimedia Wireless Networks. IEEE Communications Magazine 39 (2001) I. Katzela and M. Naghshineh: Channel Assignment Schemes for Cellular Mobile Telecommunications Systems: A Comprehensive Survey. IEEE Personal Communications Magazine, 3(3) (1996) C. Oliveria, J. B. Kim and T. Suda: An Adaptive Bandwidth Reservation Scheme for High-Speed Multimedia Wireless Networks. IEEE J. Select. Areas Communications, 16(6) (1998) J. Y. Lee, J. G. Choi, K. Park and S. Bahk: Realistic Cell-Oriented Adaptive Admission Control for QoS Support in Wireless Multimedia Networks. IEEE Transactions on Vehicular Technology, 52(3) (2003) Y. Kim, D. Lee and B. Lee: Dynamic Channel Reservation based on Mobility in Wireless ATM networks. IEEE Communications Magazine, 37(11) (1999) Z. Xu, Z. Ye, S. V. Krishnamurthy, S. K. Tripathi and M. Molle: A New Adaptive Channel Reservation Scheme for Handoff Calls in Wireless Cellular Networks. Proc. of NETWORKING 2002,Technical Committee Communications Systems of International Federation for Information Processing (IFIP-TC6),Pisa-Italy, (2002) C. Chou and K.G. Shin: Analysis of Adaptive Bandwidth Allocation in Wireless Networks with Multilevel Degradable Quality of Service. IEEE Trans. Mobile Computing 3(1) (2004). 8. C. Chou and K.G.Shin: Analysis of Combined Adaptive Bandwidth Allocation and Admission Control in Wireless Networks. Proc. IEEE Infocom,NewYork, (2002). 9. J. Wang, Q. Zeng and D. P. Agrawala: Performance Analysis of a Preemptive and Priority Reservation Handoff Scheme for Integrated Service-Based Wireless Mobile Networks. IEEE Transactions on Mobile Computing, 2(1) (2003) K. Lee and S. Kim: Optimization for Adaptive Bandwidth Reservation in Wireless Multimedia Networks. The International Journal of Computer and Telecommunications Networking 38, (2002) W. K. Lai, Y. Jin, H. W. Chen and C. Y. Pan: Channel Assignment for Initial and Handoff Calls to Improve the Call-Completion Probability. IEEE Trans. On Vehicular Tech, 52 (4), (2003) D. Lee and T. Hsueh: Bandwidth-Reservation Scheme Based on Road Information for Next-Generation Cellular Networks. IEEE Trans. On Vehicular Technology 53 (1) (2004) J. Hou and Y. Fang: Mobility-Based Call Admission Control Schemes for Wireless Mobile Networks. Wireless Communications and Mobile Computing 2001,1(3), (2001)

11 Dynamic Time-Threshold Based Scheme for Voice Calls in Cellular Networks F. Hu and N. Sharma: Priority-Determined Multiclass Handoff Scheme with Guaranteed Mobile QoS in Wireless Multimedia Networks. IEEE Trans. On Vehicular Tech., 53(1) 2004 ( ). 15. A. Olivre: Call Admission Control and Dynamic Pricing in a GSM/GPRS Cellular Network.a dissertation submitted to University of Dublin, L. Huang, S. Kumar and C. C. Kuo: Adaptive Resource Allocation for Multimedia QoS Management in Wireless Networks. IEEE Transactions on Vehicular Technology, 53(2) (2004) M. Salamah, and Idil Candan: A Novel Bandwidth Allocation Strategy for Voice Handoff Calls in Cellular Networks. Proc. of the 9 th CDMA International Conference, (2004) PTOLEMY simulation package: ETSI Standard ETR 310,, Radio Equipment and Systems; Digital Enhanced Cordless Telecommunications; Traffic Capacity and Spectrum Requirements, (1996) 20. F. Barcelo: Performance Analysis of Handoff Resource Allocation Strategies through the State-Dependent Rejection Scheme. IEEE Transactions on Wireless Communications, 3(3) (2004)