Spectral Efficiency Analysis of GSM Networks in South-South Nigeria

Transcription

1 Spectral Efficiency Analysis of GSM Networks in South-South Nigeria P. Elechi, and T.A. Alalibo Abstract n this paper, the technique of multiplicity was used to analyse GSM network capacity in Nigeria. An evaluation of Enhanced stochastic knapsack was used as an approach for resource sharing in multiservice by adopting the Erlang Loss Model in analyzing the SMS capacity. The offered traffic that is Lost Traffic based was used to dimension the system resources. To actualize this work, measurements were conducted in Benin City and Port Harcourt to determine the best signal characterization of the southern part of Nigeria. Based on the measurement data, a model was developed to predict the traffic intensity of the region. Comparisons were carried out on the different types of frequency hopping and the variant DFH based power was applied in improving the spectral efficiency of the network. The results showed that the spectral efficiency increased, as the number of cell per cluster decreased. The optimal value of the number of cell in the cluster caused reduced interference, since the reduced interference could allow the users to achieve higher rates. ndex Terms Traffic; Multiservice; Knapsack; Spectral; nterference.. NTRODUCTON Despite the numerous studies on GSM signal capacity analysis, a number of issues such as the capacity and efficiency of the network are still open. The most important among them understanding the traffic intensity of the network for proper dimensioning especially for multiservice. Traffic modelling part of networks modelling on the performance quality of any communication network. Traffic Model can be classified as smooth and non-smooth model [1] and [2]. The Smooth traffic model comprises of two types, Short Range Dependence (SRD) and Long Range Dependence (LRD) [2] [3], and [11]. A traffic model involves the following factors: Fitting nature, number of describing parameters and complexity of the parameter evaluation [4]. The introduction of Asynchronous Transfer Mode technology has strengthened the sharing of resources models with different resources requirements [6], [9] and [11]. Due to the heterogeneous nature of load traffic and the associated difficulty in its characterization; trunk reservations are used in resource control [7] and [11]. This paper is aimed at deriving a traffic model for the calculation of the number of resources that are required to handle peak Published on June 6, P. Elechi is with the Department of Electrical/Computer Engineering, Rivers State University of Science and Technology, Port Harcourt, Nigeria (tel: ; elechi.promise@ust.edu.ng). T.A. Alalibo is with the Department of Electrical/Computer Engineering, Rivers State University of Science and Technology, Port Harcourt, Nigeria. period of traffic in the cell while meeting the grade of service requirements. A single service GSM network can be dimensioned based on an associated blocking rate, the Erlang-B Law will be a suitable traffic model [9], [11], and [17]. Presently, GSM network provides different services, such as voice, data, and multimedia services [11] and [15]. Hence, Erlang-B Law can no longer be applied in traffic modelling in GSM network, rather Knapsack will be a more suitable multiservice traffic model in finding the system minimal capacity given the requested grade of service (GoS) and reproducing the resources sharing among different services and users [14], [12], and [16]. This work will examine the source, types and volume of data generation using analytical method and also investigate the kind of channel allocation with respect to the frequency hopping types to achieve the best spectral efficiency. The performance of GSM communication systems is dependent on the mobile radio channel [7] and [16]. Propagation models predict the mean signal strength and its variability at a given distance from the transmitter [9], [13], and [10]. Different models exist for different types of environments (urban and rural) [11]. Call blocking management approach will assist subscribers due to limited available radio resources, thus improving the spectral efficiency of the GSM cellular system which is of great concern to operators of GSM Network in South-South, Nigeria. There is therefore a need to tackle the problem of service congestion of GSM network.. MATERALS AND METHODS n this work, measurements were conducted to determine the traffic volume of GSM network in two major cities in south-south of Nigeria (Port Harcourt and Benin). The measurements were carried out using Samsung Mobile Phone. The traffic intensity of these locations were measured over 24hrs for 48 weeks. The average hours of the measured traffic in Erlangs were recorded. The carried traffic was expressed as the traffic load. The mean rate of blocked calls against the determined Erlang was used to find the lost traffic, to enable a formulation of the offered traffic for accurate cell dimensioning. The average number of arrival calls was evaluated with the corresponding handover arrivals. Based on the measurement data, a model was developed to analyze the network traffic intensity and spectrum. A. Multi-service Traffic Model A multi-service traffic model for GSM network was designed to meet the heterogeneous service requirements of different applications. n real application, voice calls require DO: 7

2 lower service rate than movie/picture download, therefore, movie/picture download belongs to the class that will require more allocation of channels. Many literatures have been developed on multi-server single-service using M/M/K/K system, but in this work, a multi-service loss model is developed for enhanced Knapsack Traffic model, with emphasis on the blocking probability of each class of traffic [19]. Unlike the M/M/K/K system whereby all the customers experience equal blocking probability, in this work, customer belonging to different classes experienced different blocking probabilities. Consider a GSM call with 7 channels, assuming 5 of the 7 channels are busy, if a customer that requires one channel arrives, the channel will not be blocked, but if a new arrival, that requires three channels, will be blocked. Therefore, customer that requires more channels will experience a higher blocking probability. B. Model Description Considering a set of K channels, serving customer being a member of class. Customer from class that require simultaneous Si channels and their holding times are assumed exponentially distributed with mean 1/ui. Class i customer arrival arrives according to an independent Poisson process with arrival rate, λi. The holding times are independent of each other, hence, the process of arrival and the state of the system can be expressed as: Αi = λi/ui (1) A class-i customer admitted into the channel will use Si channel for the duration of its average holding time, 1/ui, after which Si channel will be released to serve others. When a class-i customer arrives but can t find a free channel, it will be blocked and cleared from the system. The probability that an arriving class-i customer may be blocked is denoted by Bi. C. Model Derivation To develop a multiservice Traffic model for an enhanced Knapsack of capacity C, the steady state probability of process is used to satisfy the conditional probability of occurrence of state in Stochastic Knapsack [19]. Considering a set of K channels serving customers belonging to classes with traffic intensity A i each using S i resources and a knapsack of Capacity C. f j = (j i j k ) represents the possible states of the Knapsack, also consider the steady-state probability distribution of j for the case K= Let = ndependent uni-demensioned continuous-time Maxkor chamn X i (t) = Number of Class i customer in the system Where i = 1,2... Characterized by the birth rate = λ ; death rate = j i u i i (j i ) = steady state probability of process X i (t). Then i (j i ) satisfies the equations: λ i i (0) = j i u i i (2) λ i i (j i ) = j i u i i (j i + 1) for j i = 1,2,3.. (3) Normalizing (3), Substitute n = j i ji=0 i (j i ) = 1 λ = λ into (3) u = j i u i Also, in a single cell expressed as a group of S i servers serving each Class i customer. Then, According to [19] i (j i ) = where j i = 0, 1,2,3.. X i (t), i = 1,2,3.. be independent processes, Then, P(j) = P(j 1, j 2..j ) n steady state X i (t) = j 1, X 2 (t) =, j 2...X (t) is expressed as: P(j) = P (j 1, j 2,..j ) = Recall, we say j is feasible if i=1 js = i=0 j i s i k e A i (4) State are denoted with j = (j i..j N), where j i. satisfies the number of class i customers in the system. Assuming there is no waiting room and that customers who do not find sufficient resources are blocked automatically. Let P(j) = 0 for all j f to normalize, the probability j f. Recall, the probability that the infinite, server process is in a feasible state, in k finite servers C = jεf i=1 Hence, by truncating (4), the probability P(j) considered on jεf becomes P(j) = P(j 1, j 2,..j ) = 1 C i=1 (5) jεf (6) Let F(m) be the subset of an arriving class state m customer who must not be blocked, t implies that, F(m) = jεf such that = i=1 j i s i k - s m. Then, B(m)= jεf(m) P(j), m=1,2, So, using (6) B(m) = 1 - jεf(m) i=1 jεf i=1 j A i i A i j i (7) DO: 8

3 P(j) = P( i 1,i 2..i ) = 1 C i=1 C = jεf i=1 as jεf (8) Also, consider a set of K channels serving customers belonging to classes with traffic intensity A i each using S i resources and a knapsack of Capacity C. f j = ( j i, j k ) represent the possible states of the Knapsack, then the probability of occurrence of each state is denoted using (8). This equation holds by conditional probability definition. B(m) is the probability that a class m customer is blocked; in this model, B(m) is derived for each class. P(j) is the state probability vector for all jεf. F(m) represents the subset of the states for an arriving class m customer that will not be blocked. Fig. 3. llustration of blocking probability with load. RESULTS AND DSCUSSON A. Results Fig. 4. Relationship between Blocking Probability and Users Fig. 1: Average Rate of Blocked Calls in Benin City Fig. 5. The relationship between spectral efficiency with load Fig. 2. Average Handover arrival calls Fig. 6. Relationship between Spectral Efficiency and Radius DO: 9

4 Fig. 7: Spectral Efficiency against the number of cells per cluster B. Discussion n section, the stochastic knapsack model was used as an extension of the Erlang-B algorithm to find the analytical model of the blocking probability for each service of voice and data, knowing the system capacity and the incoming traffic. Fig. 1 and 2 shows the average rate of blocked calls and call handover. The result showed that the maximum blocked call occurred between 12 and 3 pm. The performance of call blocking probability is drawn versus the traffic (A) as shown in Fig. 3. Fig. 3 presents a comparison of blocking probabilities in GSM systems for (a) without FH, (b) with FH and (c) with DFH. t can be seen from the graph that the blocking probability of FH at a load of 2 Erlang is equal to This is more than GSM- without FH for the same load which is equal to 0. The blocking probability is suppressed considerably by applying DFH. f we compare the Frequency Hopping systems for a loading of 3 Erlang, GSM without FH has a blocking probability of 0, whereas with hopping is 0.08; and for DFH. As loading increases, blocking probability for GSM without FH increased very fast, while for FH, increase was considerably slow. For DFH the blocking probability was 0 from 6 Erlang load, whereas after exploiting DFH, no single user experienced outage until the load became high. n Fig. 4, the blocking probability of FH for 60 users was 0.02, which is less than GSM without FH for the same number of users at a BP of The blocking probability appeared very low in DFH and comparing between the systems for 60 to 70 users, then the GSM network without FH has an outage probability of 0.09 and decreases to with FH and 0 for DFH. As the loads increase, the blocking probability also increase very rapidly, while for FH this increase is considerably slow. For DFH the blocking probability is 0 until the number of users exceed 70. From Fig. 5 it was seen that traffic control could potentially increase the spectral efficiency by decreasing the traffic itself, since the reduced interference could allow the users to achieve higher rate. The figure showed the performance of these systems for GSM without FH, with FH, and with DFH. Also in the figure is seen the improvement in the average user spectral efficiency and in the high data rate coverage. Although for low loading, the difference in spectral efficiency was not much for 1 Erlang load, without FH, its 3.5 b/s/hz, with FH its 5.1 b/s/hz, and with DFH its 6.5 b/s/hz. The results showed that with the help of the switching relays, at a load of 2.3 Erlangs with heavy traffic. GSM network without FH will experience spectral efficiency of 1.9 bits/s/hz, with FH at 2.2 b/s/hz, and DFH at 2.9 b/s/hz. At a load of 4 Erlangs, the spectral efficiency for DFH is higher than that of GSM without FH. Fig. 6 shows the performance of Spectral efficiency with reuse distance (radius in meters). The Spectral efficiency was shown to be increasing by decreasing the reuse distance. From optimal value of spectral efficiency, the optimal value of reuse distance can be obtained, and from the optimal reuse distance also the minimum value of interference can be obtained. n Fig. 7, it is seen that the spectral efficiency increased, as the number of cell per cluster decreased. The optimal value of the number of cell in the cluster caused reduced interference, since the reduced interference could allow the users to achieve higher rates. V. CONCLUSON Based on the analysis conducted, it can be observed that the new developed variant of Dynamic Frequency Hopping has shown to be the method which does not allow received signal power to drop easily. Therefore, the received power method gives better performance than other methods of DFH. The blocking probability appeared very low in DFH and comparing between the systems for 60 to 70 users, then the GSM network without FH has an outage probability of t was observed that the spectral efficiency increased, as the number of cell per cluster decreased. The optimal value of the number of cell in the cluster caused reduced interference, since the reduced interference could allow the users to achieve higher rates. The analytical method of data for cell dimensioning and the lost traffic approach allows more accurate optimal modelling and has shown the best spectral efficiency. REFERENCES [1] A.J. Viterbi, Principles of Spread Spectrum Communication, Addison-Wesley. Pp , [2] R. Yates, (1995): A framework for Uplink Power Control. n: Cellular Radion Systems, EEE Journal on Selected Areas in Communications, vol. 13, No. 7, pp , [3] P. Elechi, mproved Voice/Data Traffic Performance of Cellular CDMA System, nternational Journal of Engineering and Technology, vol. 4, No. 7, pp , [4] S. Anand, and A. Chockalingam, Performance of Cellular CDMA with Voice/Data Traffic with an SR based Admission Control, Department of ECE, ndian nstitute of Science, Bangalore, ndia, pp.2-3, [5] G.R. Bianchi, A Novel Network Traffic Predictor based on Muthfractal Traffic characteristic, EE Communication Society, Globecom, 2004, pp , [6] J. Zander, Performance of Optimum Transmission Power Control in Cellular Radio Systems, EEE Transactions on Vehicular Technology, vol. 41, No. 1, pp , [7] J.J. Biebuma, A. Eteng, and P. Elechi, Millimeter Wave Propagation for mproved Distributed Wireless Communication System, Continental J. of nformation Technology, vol. 4, pp , [8] E. Altaman, E. et al, Optimal Call Admission Control, EEE Transactions on Communications. Pp , [9] P. Elechi, J.J. Biebuma, and P. Elagauma, Estimating CDMA Capacity and Performance in Mobile Network, nternational Journal of Engineering and Technology, vol. 3, No. 1, pp , [10] P. Elechi, and P.O. Otasowie, Path Loss Prediction Model for GSM Fixed Wireless Access, European Journal of Engineering Research and Science, vol. 1, No. 1, pp. 7-11, [11] J.J. Biebuma, and S. Orakwe, and O.J. gbekele Traffic Analysis of GSM Network in Northern Nigeria, Continental J. of nformation Technology, vol. 4, pp , DO: 10

5 [12] P. Elechi, and P.O. Otasowie, Determination of Path Loss Exponent for GSM Wireless Access in Rivers State using Building Penetration Loss, The Mediterranean journal of Electronic and Communications vol. 11, No. 1, pp , [13] B. Kuboye, Development of a Framework for Managing Congestion in GSM in Nigeria, Master Thesis, University of Maiduguri, Bornu State, [14] T.S. Rappaport, et al, Traffic Model and Performance Analysis for Cellular Mobile Network, EEE transaction on Vehicular Technology, pp , [15] J.S. Kaufman, Blocking in a Shared Resources Environment, EEE Transaction On Communications, pp , [16] Cisco nc., Traffic Analysis: Visited 24/12/2015 [17] A. K. Erlang, Solution of Same Problems in the Theory of Probability of Significance in Automatic Telephone Ex. The Post Office Engineers Journal, vol. 10, pp , [18] P. Taylor, nsensitivity in Stochastic Models : Chapter 3, in N. Van Dijk and R. Boucherie (eds.): Queueing Networks: A Fundamental Approach, Springer-Verlag, Berlin, New York: , [19] M. Zukerman, ntroduction to Queueing Theory and Stochastic Teletraffic Models, Electrical Engineering Department, City University of Hong Kong, , Elechi Promise received his B.Eng and M.Eng degrees in Electrical/Electronic Engineering from University of Port Harcourt, Choba, Rivers State, Nigeria in 2006 and 2011 respectively. He is currently a Ph.D Student in Electronic and Telecommunication Engineering of University of Benin. He is a corporate member of the Nigerian Society of Engineers (NSE), nternational Association of Engineers (AEng), nstitute of Engineering and Technology (ET), and Nigerian nstitution of Electrical/Electronic Engineers (NEEE) and also a registered practicing Engineer with the Council for Regulation of Engineering in Nigeria (COREN). His current research interest is on Radio Propagation for Mobile communications, GSM Technology, Microwave Propagation, Signal Analysis, NanoTechnology and CT. He is currently a Lecturer in the Department of Electrical and Computer Engineering, Rivers State University of Science and Technology, Port Harcourt, Nigeria. DO: 11