Improvement in the Priority Handoff Scheme for Multi-Service Wireless Mobile Networks

Transcription

1 Research Journal of Applied Sciences, Engineering and Technolog 6(2): , 213 ISSN: ; e-issn: Maxwell Scientific Organization, 213 Submitted: Januar 17, 213 Accepted: Februar 22, 213 Published: November 1, 213 Improvement in the Priorit Handoff Scheme for Multi-Service Wireless Mobile Networks 1 Cheng-Gang Liu, 1 Zhen-Hong Jia, 2 Xi-Zhong Qin, 2 Lei Sheng and 2 Li Chen 1 School of Information Science and Engineering, Xinjiang Universit, Urumqi 8346, China 2 Subsidiar Compan of China Mobile in Xinjiang, Urumqi 8363, China Abstract: In this stud, a new handoff strateg to improve the performance of wireless mobile networks is presented. It has been found form this stud that the dropping probabilit of handoff calls is drasticall reduced compared to the existing method of channel reservation strateg of handoff calls and the performance of new calls can be improved. The strateg is that the new call should be delaed then it can use the last idle channel. The time that all the channels are holding has been shortened. Opportunities of the new call and handoff call occupanc channel are increased; the scheme also takes into account the priorit of different data tpes and onl when the highpriorit data packets queue is empt, the data packets in low-priorit data queue can be transmitted. Simulation results show that, to let the new voice call dela in the allowable range, can effectivel reduce the dropping probabilit of handoff call and the blocking probabilit of high-priorit data while improving the probabilit of new call to enter into the sstems. Kewords: Handoff, Markova process, multi-service, priorit, queuing theor INTRODUCTION With the rapid development of mobile communication technolog, support for multimedia applications in an increasing number of users has forced service providers to emplo cells of smaller size so that availabilit of adequate capacit could be ascertained. Meanwhile, due to accelerate the moving speed of the users, the numbers of the users that need to handoff and the handoff during an ongoing call are increased continuousl. Handoff failure that results in the forced termination of an ongoing call will add. Besides, the growth of the transmission bandwidth expanded the tpes of mobile communications service. In addition to voice services, data services such as Internet, image also appeared. In order to effectivel use wireless resources and improve the sstem capacit and service QoS, designed handoff strateg that is suitable for voice, data and multi-service has been a hot research (Hamad et al., 211; Sharna et al., 211; Yuan-Yuan et al., 21). The grade of service of a wireless mobile network can be defined in terms of the blocking probabilit of new call and the dropping probabilit of handoff call.in handoff strateg, in order to improve the two indicators, two priorities are considered in man handoff strategies. Some handoff strateg taking into account different priorities between different service tpes have been proposed (Krishna et al., 29; Al Khanjari et al., 211; Suleiman et al., 211), the different priorities between handoff calls and new calls other handoff are considered in other strategies. A new call being blocked is not as disastrous as a handoff call being dropped, so these strategies (Idil and Muhammed, 27; Sheu et al., 27; Nasr et al., 21) consist of giving higher priorit to handoff call than new call. At the same time, according to the different tpe of data service, consideration should also be given priorit of different data tpes. The (Krishna et al., 29) proposed a movable boundar strateg; real time service has a higher priorit than the non-real-time service. In order to increase the bandwidth utilization, non-real-time service is allowed to borrow unused channels from the real time service. Since handoff calls and new calls use the same set of channels, the dropping probabilit of handoff call is high when there are too man new calls. The (Idil and Muhammed, 27) proposed the guard channel strateg that gives the handoff call a higher priorit than a new call. This strateg consists of having a fixed number of channels in each cell, reserved exclusivel for handoff call. It effectivel reduces the dropping probabilit of handoff call. However, the drawback for using guard channel for handoff call is that the channels that are reserved ma be unoccupied for a long duration and new calls are blocked because the have no channels can be used. The (Haw-Yun and Jean-Lien, 24) proposed the guard channel strateg that consists of having a fixed number of channels in each cell, reserved exclusivel for data call. Corresponding Author: Zhen-Hong Jia, School of Information Science and Engineering, Xinjiang Universit, Urumqi 8346, China 3838

2 When the arrival rate of voice services is high, voice services cannot occup the guard channel of data services. This will result in a high blocking probabilit of voice service. The (Zhihua, 28) have recentl proposed a strateg to dnamicall reserve guard channel for handoff calls. This increased the bandwidth utilization, but did not consider the different priorities between the various data services. The (Huang and Chan, 24) proposed a movable boundar and guard channel strateg, the consider two priorities that real time service has a higher priorit than the non-real-time service and the handoff call has a higher priorit than a new call. The dropping probabilit of handoff call is significantl lower. However, the performance of new calls has been reduced and this strateg did not consider the different priorities between the various data services. In the present stud, a new handoff strateg based on multi-service is proposed. When services are bus in wireless mobile networks, handoff calls have a higher priorit than new calls when call request. Idle channel can be assigned to handoff call immediatel. If there is onl one idle channel in the cell, new call would be delaed before it occup the idle channel. Although the dela time of new calls are increased in new strateg, the probabilit of new call and handoff call to obtain a free channel are increased, the qualit of voice service are improved. In a data packet communication environment, data packets have different requirements. Some non realtime data packets can tolerate delas whereas real-time data packets are sensitive to delas. At the same time the strateg to provide differentiated services for data service can create more value for the communication operators. Combined with these characteristics of current data service in wireless mobile networks, new strateg takes different queue policies. According to the different importance of data services, data services are divided into high and low priorit packet. For the smooth presentation of the idea of the new handoff strateg for performance improvement, the 4D Erlang theor with new handoff strateg is described. The reserved channel strateg and movable boundar channel strateg are compared with the new handoff strateg. SYSTEM MODEL The sstem model is shown in Fig. 1. In the following, we assume that the total number of channels in the channel pool shared b voice service and data service in each cell is fixed and denoted b C.N channels can be allocated to voice services and C-N channels are allocated to data services. The voice calls are divided into handoff calls and new calls. Voice calls onl can occup the previous N channels. Handoff calls Res. J. Appl. Sci. Eng. Technol., 6(2): , Fig. 1: Sstem model and new calls are endowed with the same priorit in the previous (N-1) channels. Handoff calls are assigned priorit over new calls when the N-idle channel is allocated to voice call. When the N-th channel is idle in the cell, if the handoff call arrives, the N-th channel can be directl occupied b the handoff call; if the new call arrives, the N-th channel cannot be directl occupied b the new call, the new call would be delaed. During the dela time, if the users in the previous (N-1) channels complete the call, then the idle channel in the previous (N-1) channels can be occupied b the new call. If there is still no idle channel when the dela time is over, the N-th idle channel can be occupied b new call. According to the different importance of data services, data services are divided into high and low priorit packet. When a data call arrives, if it is highpriorit packet. It is stored in the queue Q 2. If it is a low-priorit packet. It is stored in the queue Q 1. When the data packet is transmitted, the packet is removed from the queue. When the queue Q 2 is empt, the low priorit packets can be transmitted in the queue Q 1. When the channel (C-N) of the data service is bus, the channel assigned to the voice service in the idle time can be occupied b data services. When a voice call arrives, an idle channel from N channels can be assigned to the call. If there is no free channel can be occupied b voice call, while channels of the voice service are occupied b data services, the channel must be released immediatel b the data call, the data call is placed in the queue Q 2 until there are the idle channels can be occupied. If the previous (N-1) channels are occupied b all the voice calls, the voice call is treated differentl according to the tpe of the call. In Fig. 1, The N-th channel can be occupied b handoff call immediatel. The new call would be delaed. During the dela time, the sstem has been in a state D. In the D state, there is at least one idle channel in the cell. The time that all channels are

3 occupied is reduced. The time that the cell stas in the (N-1) state has been extended. Thereb the dropping probabilit of handoff call and the blocking probabilit of new call are reduced. When the dela time d (t) is over, if the users in the previous (N-1) channels leave the cell, then the idle channel in the previous (N-1) channels can be occupied b the new call. If there is not idle channel to be occupied in the previous (N-1) channels, the N-th channel can be occupied b the new call. TRAFFIC MODEL This queuing sstem with C servers and an infinite number of users consisting of both voice and data service accords to Poisson process. The arrival rate of new calls is λ o ; the arrival rate of handoff calls is λ h. Their service times are negative exponential distribution (Rivero-Angeles et al., 29). Their service rate are μ o and μ h. The maximum length of the data packets waiting queue is Q 1 and Q 2 (Q 1 >Q 2 ), the arrangement rule is FIFO. The arrival rate of low priorit packets and the arrival rate of high priorit packets were λ d1 and λ d2. The length of data packet is negative exponential distribution, the mean length of data packet is L d bits. Assuming that the transmission of data packet uses the same encoding, the transmission rate of data packet is V d kbit/s in the coding mode. The average transmission time of the low priorit data packet and the high priorit data packet is 1/μ d = L d /V d. We represent the state of the cell with a fourdimensional Markov chain with states (i, j,l, g), where i number of channels used b new calls in N; j number of channels used b handoff calls in N; l number of high priorit data calls in Q 2 ; g number of data calls in cell. The value ranges of discrete parameters are i [, N], j [, N], l [, Q 2 ], g [, C + Q 1 + Q 2 ]. Since the stead-state distribution of the multidimensional continuous time Markov process is not solved easil and the actual sstem meets μ d / (μ o + μ h ) >>1, the four-dimensional continuous time Markov process is decomposed into two consecutive twodimensional continuous time Markov process. The two parts are voice calls and data calls (Ghani and Schwartz, 1994). Voice calls section: The state transition diagram of voice call is shown in Fig. 2. For i and j of State (i, j), i is the number of new calls in the cell and j is the number of handoff calls in the cell. δ -1 = d, d is the dela time of new call. State D represents the state that the new call is delaed; the arrival rate to reach the state D is δ. If a whole row is considered at the -th position of a column, then the stationar probabilit states are given b the following equations: Res. J. Appl. Sci. Eng. Technol., 6(2): , (1), 1, 1,, 2 2, (2) x, 1, x 1 x x x1, (3) where, π (x, ) indicates the stationar probabilit of the state (x, ). Let ρ o = λ o /μ o, ρ o is the call intensit of new call. Then we can think of the arrival of a new call is Poisson distribution parameters for ρ o. The service time is negative exponential distribution parameters for 1. Stead-state probabilit for each row of N states (x, ) is: x,, x n, n 1 (4) x x! If a whole column is considered at the x-th position of a row, Let ρ h = λ h /μ h, ρ h is the call intensit of handoff call. Then we can think of the arrival of a handoff call is Poisson distribution parameters for ρ h. The service time is negative exponential distribution parameters for 1. Stead-state probabilit for each column of (N+1) states (x, ) is: h,, x n, n 1 (5) x x! B substituting x = in formula 5 and the resulting π (,) so obtained, if substituted in formula 4, results in the following expression for the stationar probabilit states π (x,) : x h,, x n, n 1 (6) x x!! In addition to (6), the following equations are derived for finding the remaining states π (D,) and: π (n-, ) : / n 1 1, n n2, n 1 D, n n, ( n 1) D, n 1, (7) (8) n n, (9) D, After solving the above equations, the following relations, the expressions for π (D, ) and π (n-, ) are obtained:

4 Res. J. Appl. Sci. Eng. Technol., 6(2): , 213 Fig. 2: State transition diagram of voice call Fig. 3: State transition diagram of data call n1 h ( n 1)! n 1! D,, (1) Y g g! 1!!! n1 n1 n1 n h n n h (13) ( n 1)! n! n h, (11) n, g ( n 1) (14) where, The dropping probabilit of handoff call in this case is:, x h h x! x! n 1 n Y 1 (12) P d n n n! n! h g, (15) n, 3841

5 And the blocking probabilit of new call is obtained as: P n1 n b D, n, n1 n1 n n h h g,,! n 1!! n! Res. J. Appl. Sci. Eng. Technol., 6(2): , 213 (16) Data calls section: When there are w (w = i + j) channels occupied b voice call in the sstem, the maximum of data calls simultaneousl serviced is k (k = C-w) in the sstem. For l and g of state (l, g), l is the number of the high priorit packets in the queue Q 2, g is the total number of packets in the sstem. The state transition diagram of data call is shown in Fig. 3. Each state transition diagram of data call is described in Fig. 3, the principle of which is the flow conservation. Let inflow speed is equal to the outflow rate for each state. We can get the stead state balance equation: ( ) ( k1) u gkl, l (17) d1 d2 ( l, g) d lg, 1 ([1 u( g k Q )] [1 u( l Q )] 1 d1 2 d 2 u( l 1) ku ( l) ( g k 1) ku ) d d l, g ul ( 1) d1 1, d2 1, 1 l g l l g g k d2 l, g1 uq l ku + ( 1) 2 d l1, g1 l u( k Q1 g 1) ku d l, g 1 k g k Q Q, l Q (18) where, π (l, g) is the state transition probabilit of the sstem that is in state (l, g). Let u (x) denote the step function, which is defined as follows: u (x) = 1 when x and u (x) = when x<; (x) = 1 when x = and (x) = when x. For state equation of formula (15) and the formula (16), there are (k + (Q 1 + 1) (Q 2 + 1)) equations. We coupled on the state probabilit normalized condition as formula (19); we can obtain the stead-state solution of state b solving linear equation: k 1 Q k Q Q (, g ) l,g g l g k l 1 (19) The dropping probabilit of low priorit data call in this case is: Q 2 d1 ( l, g Q1 l) l P (2) 3842 And the dropping probabilit of high priorit data call is obtained as: P k+q 1+Q2 d2 Q 2, g g= k+q 2 SIMULATION RESULTS AND DISCUSSION (21) For the numerical appreciation of the obtained results, the following parameters are assumed for a cell of a mobile cellular sstem: number of channels in the cell, C = 14, number of channels for voice services N = 1, high priorit queue length Q 1 = 2, low priorit queue length Q 2 = 1, average length of data packets, L d = 12 bit. Assuming that the transmission of data packet uses the same encoding, the transmission rate of data packet is V d = 9.5 kbit/s in the coding mode. The arrival rate of new calls λ o = ( ) /min, the arrival rate of handoff calls λ h = ( ) /min. Their service rate are μ o =.67/min and μ h =.83/min. The arrival rate of low priorit packets was λ d1, the arrival rate of high priorit packets was λ d2. λ d1 = 2λ d2, λ d1 =.45 packets per second. Using MATLAB simulation software, when change the value of λ, λ h, we can get the blocking probabilit of new call is P b, the dropping probabilit of handoff call is P d, the dropping probabilit of low priorit data call is P d1, the dropping probabilit of high priorit data call is P d2. Reserved channels for handoff calls, r = 3 in the reserved channel strateg, number of channels for data calls in the cell, m = 4. In the movable boundar channel strateg, the channel is not reserved for the handoff calls. When the voice dedicated channel is idle, the data call can occup the voice channel, but when a voice user arrives, the data call must be released immediatel channel. For ease of comparison, the new switching strategies proposed in this stud and the protection of channel strateg and movable boundar polic, all consider the priorit of different data tpes. In Fig. 4 and 5, since new call would be delaed before it occup the idle channel. The time of all channels were occupied is shorten, the time of the sstem in the state (N-1) has been extended. The probabilit of new call and handoff call to obtain a free channel are increased. Compared with reserved channel strateg and movable boundar strateg, the dropping probabilit of handoff calls P d has been greatl improved in the new handoff strateg. In Fig. 6 and 7, since there are not reserved channels in the new handoff strateg and movable boundar strateg, the new calls are less affected b handoff calls, there are smaller blocking rate in these two strategies than reserved strateg. The probabilit of an idle channel after a dela of a new call in the new

6 Res. J. Appl. Sci. Eng. Technol., 6(2): , 213 Fig. 4: The dropping probabilit of handoff call b arrival rates of the new call Fig. 5: The dropping probabilit of handoff call b arrival rates of the handoff call Fig. 6: The blocking probabilit of new call b arrival rates of the new call 3843

7 Res. J. Appl. Sci. Eng. Technol., 6(2): , 213 Fig. 7: The blocking probabilit of new call b arrival rates of the handoff call Fig. 8: The dropping probabilit of handoff call b the dela time of new call Fig. 9: The blocking probabilit of new call b the dela time of new call 3844

8 Res. J. Appl. Sci. Eng. Technol., 6(2): , 213 Fig. 1: The dropping probabilit of data packet b arrival rates of the new call Fig. 11: The dropping probabilit of data packet b arrival rates of the handoff call handoff strateg is greater than the movable boundar strateg, so the blocking probabilit of new call in the new handoff strateg is small than the movable boundar strateg. In Fig. 8 and 9, the arrival rate of new calls λ o = 5.358/min, the arrival rate of handoff calls λ h = 4.14/min. Their service rate are μ o =.67/min and μ h =.83/min. With the growth of the dela time of a new call, the probabilit of the idle channel will increase when the call arrives. Since the handoff calls are gave higher priorit, an idle channel can be occupied b the handoff call immediatel, the dropping probabilit of handoff calls decreased more significantl. Since the N- the idle channel is not allowed b the new calls, the blocking probabilit of 3845 new call is not changed significantl as the dela time increases. In Fig. 1 and 11, the new handoff strateg is proposed that onl when the high-priorit data packets queue is empt, the data packets in low-priorit data queue can be transmitted. Compared with the dropping probabilit of low priorit data call P d1, the dropping probabilit of high priorit data call P d2 has been greatl improved. Meanwhile, since the new call is not able to occup the channel reserved for the handoff calls and data calls can borrow reserved channels in the reserved channels strateg, the dropping probabilit of data calls in the reserved channels strateg should be less than the new handoff strateg. However, the dropping probabilit of

9 Res. J. Appl. Sci. Eng. Technol., 6(2): , 213 data calls will rapidl increase when the handoff call arrival rate is high. CONCLUSION A new handoff strateg based on business was presented in this stud. In the actual communication process, users are more willing to accept occupies channel after the appropriate dela not directl blocked. A new call was allowed occuping idle channel after dela within the scope in this polic. The probabilit of the voice call into the sstem was increased. And the dropping probabilit of handoff call was reduced. The qualit of service of a new call was improved at the same time. The dropping probabilit of high-priorit data was reduced because it gives the data call a different priorit. In order to reflect actual problems, we consider that the service time of handoff calls and new calls are different. New handoff strateg for improving network performance has certain significance. ACKNOWLEDGMENT This stud is supported b 212 Research and Development Fund Project of China Mobile Communications Group Co., Ltd. Xinjiang (Grant No. XJM212-1). REFERENCES Al Khanjari, S., B. Arafeh, K. Da and N. Alzeidi, 211. An adaptive bandwidth borrowing-based call admission control scheme for multi-class service wireless cellular networks. Proceeding of the International Conference on Innovationsin Information Technolog (IIT), Abu Dhabi, pp: Ghani, S. and M. Schwartz, A decomposition approximation for the analsis of voice/data integration. IEEE T. Commun., 42(7): Hamad, A., M., Ehab and S. Adel, 211. Performance analsis of a handoff scheme for two-tier cellular CDMA networks. Egpt. Inform. J., 12(2): Haw-Yun, S. and C.W. Jean-Lien, 24. The stud of dnamic multi-channel scheme with channel deallocation in wireless networks. Comput. Netw., 45(4): Huang, Q. and S. Chan, 24. An enhanced handoff control scheme for multimedia traffic in cellular networks. IEEE Comm. Lett., 8(3): Idil, C. and S. Muhammed, 27. Analtical modeling of a time-threshold based bandwidth allocation scheme for cellular networks. Comput. Commun., 3(5): Krishna, P.V., S. Misra, M.S. Obaidat and V. Saritha, 29. An efficient approach for distributed dnamic channel allocation with queues for realtime and non-real-time traffic in cellular networks. J. Sst. Softw., 82(7): Nasr, A.A., H.M. El-Badaw and H.M. El-Hennaw, 21. Connection admission control reward optimization for different priorit classes in homogeneous wireless network. Proceeding of the 6th International Conference on Wireless and Mobile Communications (ICWMC), Valencia, 21: Rivero-Angeles, M.E., D. Lara-Rodriguez and F.A. Cruz-Perez, 29. Differentiated back off strategies for prioritized random access dela in multiservice cellular networks. IEEE T. Veh. Technol., 58(1): Sharna, S.A., M.R. Amin and M. Murshed, 211. Call admission control polic for multiclass traffic in heterogeneous wireless networks. Proceeding of the 11th International Smposium on Communications and Information Technologies (ISCIT), Hangzhou, 211: Sheu, T.L., Y.J. Wu and B. Li, 27. A generalized channel preemption model for multiclass traffic in mobile wireless networks. IEEE T. Veh. Technol., 56(5): Suleiman, K.H., H.A. Chan and M.E. Dlodlo, 211. Analsis of serving discipline algorithms for cellular networks. Proceeding of the 2nd International Conference on Wireless Communication, Vehicular Technolog, Information Theor and Aerospace & Electronic Sstems Technolog (Wireless VITAE), Chennai, 211: 1-6. Yuan-Yuan, S., W. Ren-Yong, L. Ren-Fa, D. Wei and L. Jian-Hui, 21. A novel measurement-based call admission control algorithm for wireless mobilit networks under practical mobilit model. Proceeding of the International Conference on Communications and Mobile Computing (CMC), Shenzhen, 3: Zhihua, Z., 28. Performance analsis of a QoS guaranteed dnamic channel reservation for handoff prioritization in cellular mobile networks. Proceeding of the 4th International Conference on Wireless Communications (WICOM 8), USA, 28: