International Conference on Information and Communication Technology and Its Applications (ICTA 2016) Federal University of Technology, Minna, Nigeria November 28 – 30, 2016 Traffic Congestion Analysis in Mobile Macrocells Aliyu Ozovehe1, Okpo U. Okereke2, Anene E. C.2, and Abraham U. Usman3 1 Trifield Technology Limited, Abuja, Nigeria 2 Department of Electrical and Electronics Engineering, Abubakar Tafawa Balewa University, Bauchi, Nigeria 3 Department of Electrical and Electronics Engineering, Federal University of Technology, Minna, Nigeria 1 aliyu123oz@gmail.com Abstract—Traffic congestion during busy hour (BH) pattern of busy hour traffic is considered for congestion deteriorates the overall performance of cellular network and evaluation [5] using key performance indicators (KPIs). may become unmanageable unless effective and efficient These KPIs are defined for different interfaces and methods of congestion control are developed through real live network elements of GSM/GPRS. To cater for subscriber traffic data measurement and analysis. In this work, real live demand, radio frequency (RF) optimization teams use the traffic data from integrated GSM/GPRS network is used for KPIs to generate quality of service (QoS) reports to ensure traffic congestion analysis. The analysis was carried out on 10 minimal congestion over all the interfaces and network congesting cells using network management system (NMS) elements in order to avoid QoS degradation by maintaining statistics data that span three years period. Correlation test the KPIs under pre-defined threshold [6]. was used to show that TCH congestion depend only on call setup success rate (CSSR) and BH traffic at cell level. An Network congestion leads to poor QoS which affect average correlation coefficient value of 0.9 was observed grade of service (GoS) of the network, particularly during the between TCH congestion and CSSR while 0.6 was observed busy hour of the day [7]. In a loss system, the GoS describes between TCH congestion and BH traffic. The correlation test is that proportion of calls that are lost due to congestion in the important when selecting input for congestion prediction busy hour and can be measured using equation (1): modeling. Number _ of _ lost _ calls Keywords-traffic congestion; GSM; GPRS; macrocells GoS  (1) Number _ of _ offered _ calls I. INTRODUCTION While a whole range of services of GSM technology are All over the world, cellular network operators are faced in use in Nigeria, it is obvious that the network performance with the challenges of improving the quality of service (QoS) in terms of QoS are degrading which proved that GSM while increasing capacity and rolling out new services. This network is either over utilized or under dimensioned. Hence has resulted in many congested networks and consequently the need for this analysis to identify the cells that are degradation of QoS due to inadequate provision of the responsible for congestion during busy hour by statistically needed resources or underutilisation of the available analyzing traffic data of a live network in order to establish resources. the presence or absence of congestion. To cope with subscriber demands and meet Regulator standards, cellular network providers dimension network elements on continuous based using network management II. LITERATURE REVIEW system (NMS) statistics, drive test trailing and customer feedbacks. Traffic analysis is important for evaluating the However, Nigerian Communication Commission (NCC) performance of existing networks and network optimization. quarterly audit reports of GSM network performance had The events that occur in BTS trigger different counters in the consistently shown that the operators have not been able to BSC and MSC memory that are used for performance meet the set standards [1] due to network congestion [2]. If monitoring and network evaluation. Various KPIs that are there is no hardware fault, network congestion usually occurs used to measure network performances are derived with the due to insufficient number of radio channels in access help of these counters using different formulations [7]. In network elements [3]. practice, the performance can be monitored at different This work used busy hour traffic data of access network sections of the network [8] and in this work the network from a live network to analyse traffic congestion in some performance is evaluated at cell levels in terms of resource macrocells of GSM/GPRS network. The busy hour of a allocation and resource utilization. network is the time when the network processes the Some of the early works on GSM network elements highest traffic in a day and it is used to measure network performance were done mostly on access part of the network performance, determine the robustness of a network and its particularly at BTS level. For example, [9] proposed a traffic dimension [4]. To measure the network performance, the model for mobile network, using Markov chain to determine 243 International Conference on Information and Communication Technology and Its Applications (ICTA 2016) call blocking and handoff failure probabilities when no standalone computer and exported to Microsoft Excel queuing of new or handover calls is performed while [10] environment. modeled the effect of user mobility on the performance of The network is composed of 742 cells, 262 BTSs. mobile networks using office hours traffic data. Location Measurements were taken from November, 2012 to updates were analyzed to evaluate the probability of call September, 2014. Correlation test was used to choose only dropping when handover is needed. Likewise, [11] analyzed KPIs that have significant effect on TCH congestion during seventy eight traffic channels and showed that a single busy hour. dedicated channel is enough for good QoS. [12] investigated Correlation coefficient is defined as a number or function GSM/GPRS system performance using dedicated number of indicating the degree of correlation between two variables GPRS channels and some idle periods between voice calls like X and Y. In this work, the variables are busy hour traffic for GPRS data packet transfers. Reservation of more and CSSR, HOSR, DCR, SDCCH and TCH Congestions as channels brings handover failure and dropped call KPIs to measure the network performance. Equation (2) probability to very small values but lack of ordinary channels defines correlation coefficient as: produces a larger new calls blocking probability. The work of [13] presented the results of study of DCS1800 Um-interface using daily traffic measurement Correl ( X , Y )   x  x  y  y  (2) data. The performance indicators used are Traffic, CSSR,  x  x    y  y  2 2 handover success rate (HOSR), standalone dedicated control channel (SDCCH) and traffic channel (TCH) congestion. The analysis shows the limitations of the system to Microsoft excel statistical tools was used for the accommodate extreme offered traffic without adding more correlation analysis. resources. A model combining simulations for paging, signaling and traffic channels was developed by [14] to investigate the optimal dimensioning of a finite physical IV. BTS DAILY BUSY HOUR TRAFFIC ANALYSIS resource allocated across multiple logical channels with multiple traffic types. The busy hour traffic data showed that 154 cells Reference [15] evaluated the performance of GSM1900 experienced congestion out of 742 cells during the period Um-interface of two different vendors using daily under investigation. In the cell KPIs analysis, 10 worst measurement data for one week. The performance indicators congesting cells were chosen for BH TCH congestion used are peak hour traffic; CSSR; Handover Failure; analysis [19].The ten most congested cells is shown in Table congestion on control channels (SDCCH blocking rate); 1, Table 2 presents the maximum and average daily BH congestion on traffic channels (TCH blocking Rate); drop on traffic of the ten most congested cells over the period. traffic channels; drop on control channels; cells with TCH congestion rate exceeding 2% and TCH Mean Holding Time. V. BTS TCH CONGESTION CORRELATION TEST ANALYSIS In another work, [3] analyzed traffic data from a trunked Using the BTS KPIs and BH traffic, the correlation test radio network operated by Ecomm in UK using OPNET showed that the ten worst cells behaved differently in terms model. The findings indicated that traditional Erlang models of KPIs that have strong correlation with TCH congestion for voice traffic may not be suitable for evaluating the during busy hour which implied that they must be performance of trunked radio networks. In a related work, investigated differently based on their correlation results. [16] formulated a dynamic channel allocation model using However, we selected 4 cells to know which KPI has strong Markov chain technique as an improvement on [17]. There is correlation with TCH congestion at cell level. The result of one problem common to all these works at BTS level, the test for the four cells is summarised in Table 3. exclusive handover channels were employed for easy Table 3 shows that BH TCH congestion has strong compliance of QoS standards for ongoing calls and handover correlation with CSSR and busy hour traffic for the four failure reduction. However, the disadvantage is that new cells. calls blocking increase as a result of the reduction of Given the result of the correlation test, the maximum available ordinary channels. The solution should have been daily BH traffic carried by the ten cells and it impact on that resources should be added to maintain GoS of the CSSR and TCH congestion is shown in Figure 2 and Figure network as put forward by [18] and supported by the work of 3. [13]. From Figure 2 and 3, most of the cells CSSR and TCH CONG degraded when they carried maximum traffic while III. METHODOLOGY Cell 396B, 393C and 301C suffered worst KPIs degradation. The setup for data collection in GSM/GPRS network The behaviour of the ten worst cells when they comprises of base station subsystem (BSS) and network experienced worst TCH CONG and the effect on other KPIs subsystem (NSS) connected to standalone system called during the period is shown in Figure 4 to 6. NMS as shown in Figure 1. All the cells CSSR degraded and their ability to carry NMS is the functional entity from which the service traffic is limited when they experienced worst TCH CONG provider monitors and controls the entire network. The data while Cell 496B, 430C, 396B and 038B recorded worst KPIs used in this work was extracted from the NMS with the help degradation. This shows that traffic channel of these cells are of Ericsson Business intelligent (BI) tools installed on the not properly dimensioned. 244 International Conference on Information and Communication Technology and Its Applications (ICTA 2016) VI. CONCLUSION BH traffic has an average correlation coefficient of 0.9 and The analysis of traffic channels in an network existing 0.6 respectively for the cells. The strong correlation showed showed that TCH congestion beyond the acceptable 2% that the knowledge of CSSR and busy traffic can be used to threshold for traffic channel (TCH) occurred in 154 cells out predict TCH congestion which is crucial for cellular network of 742 cells investigated. optimization and resource management. The busy hour TCH congestion analysis showed that the congestion depends on CSSR and BH traffic. The CSSR and Figure 1. System for collecting traffic data TABLE I. TEN WORST CONGESTING CELLS Cell Ids Dates FC0396B FC0301C FC0362C FC0385B FC0393C FC0430C FC0494B FC0496B FC0038B FC0725C Jan-13 3.98 60.18 80.01 2.14 4.35 78.95 46.04 69.59 3.7 - Feb-13 14.44 46.81 23.86 27.91 13.83 24.64 19.92 - 14.67 - Mar-13 50.87 3.96 62.71 - - - - 20.4 2.41 44.16 Apr-13 20.78 30.47 50.41 32.12 19.74 24.28 55.9 - 7.53 - Jun-13 15.31 48.24 59.02 9.39 21.63 28.87 33.77 10.23 2.39 - Jul-13 10.02 53.61 69.77 4.11 14.72 13.31 17.12 7.65 - 30.85 Aug-13 7.15 37.83 67.51 2.17 18.28 18.13 16.78 3.19 - 38.36 Sep-13 10.18 33.95 65.27 - 19.57 5.26 14.01 1.86 - 16.22 Oct-13 9.99 52.04 61.52 3.1 21.07 12.43 21.69 6.05 16.97 2.76 Nov-13 13.8 58.28 63.61 3.03 11.34 3.68 36.6 4.32 - - Dec-13 3.42 74.94 84.54 13.31 18.45 9.28 31.63 32.65 - - Feb-14 16.65 71.21 80.59 14.57 9.69 82.61 54.85 83.13 17.22 19.15 245 International Conference on Information and Communication Technology and Its Applications (ICTA 2016) TABLE II. DAILY BH TRAFFIC (ERLANG) FOR THE TEN WORST TABLE III. SUMMARY OF TCH CONGESTION CORRELATION TEST CELLS TCH Correlation Cell Max Traffic Average Traffic C H SDC FC0725C 25.6 15.01 S D O CH S C S CON Tra FC0496B 43.93 18.11 Cell R R R G ffic 0. 0. 0. FC0494B 21.68 13.61 FC0725C 9 2 1 0.4 0.5 1. 0. 0. FC0430C 64.03 37.9 FC0393C 0 1 4 0.1 0.6 FC0396B 21.59 12.75 0. 0. 0. FC0362C 9 7 1 0.0 0.7 FC0393C 19.82 13.85 0. 0. 0. FC0301C 7 6 1 0.4 0.6 FC0385B 47.65 13.08 0. 0. 0. FC0362C 27.66 15.77 Average 9 4 2 0.2 0.6 FC0301C 34.99 17.12 FC0038B 54.41 21.72 100.00 98 98 90.00 80.00 70.00 CELL KPI [%] 60.00 50.00 40.00 30.00 20.00 10.00 0.00 CSSR HOSR FC0725C FC0496B FC0494B FC0430C FC0396B FC0393C FC0385B FC0362C FC0301C FC0038B NCC Figure 2. CSSR and HOSR for ten worst Cells during Max. Traffic 246 International Conference on Information and Communication Technology and Its Applications (ICTA 2016) 50.00 40.00 CELL KPI 30.00 20.00 [%] 10.00 2 2 0.20 0.00 DCR SDCCH CONG TCH CONG FC0725C FC0496B FC0494B FC0430C FC0396B FC0393C FC0385B FC0362C FC0301C FC0038B NCC Figure 3. DCR, SDCCH CONG and TCH CONG for ten worst Cells during Max. Traffic 120.00 98 98 100.00 80.00 CELL KPI [%] 60.00 40.00 20.00 0.00 CSSR HOSR FC0725C FC0496B FC0494B FC0430C FC0396B FC0393C FC0385B FC0362C FC0301C FC0038B NCC Figure 4. CSSR and HOSR for the Cells during Worst TCH CONG 247 International Conference on Information and Communication Technology and Its Applications (ICTA 2016) 35.00 30.00 CELL KPI [%] 25.00 20.00 15.00 10.00 5.00 2 0.20 0.00 DCR SDCCH CONG FC0725C FC0496B FC0494B FC0430C FC0396B FC0393C FC0385B FC0362C FC0301C FC0038B NCC Figure 5. DCR and SDCCH CONG for the Cells during Worst TCH 40.00 35.00 30.00 CELL TRAFFIC [ERLANG] 25.00 20.00 15.00 10.00 5.00 0.00 TRAFFICC_WORST TCH CONG TRAFFIC_AVERAGE Figure 6. Traffic Average and Traffic Carried by the Cell worst TCH CONG 248 International Conference on Information and Communication Technology and Its Applications (ICTA 2016) [10] Y. C. Min, S. K. Tai, S. C. Ho and Dan, K. S. Modeling of Terminal REFERENCES Mobility to Evaluate the Number of Location Updates, 1997 IEEE [1] Nigerian Communications commission Website, Retrieved October International Conference on Volume 3, Issue , 8-12 Jun 1997 25, 2015, from: Page(s):1266 – 1270 http://www.ncc.gov.ng/index.php?option=com_content&view=article [11] A. M. Francisco and Luis, M. C., Mobility Effects on Teletraffic in &id=332&Itemid=104 GSM, Instituto de Telecomunicações / Instituto Superior Técnico, [2] O. Aliyu, Okereke O.U. and Anene E.C, Literature Survey of Traffic Technical University of Lisbon, 1999 Analysis and Congestion Modeling In Mobile Network, IOSR [12] A. Hugo, C. José and M. C. Luis, Analysis of a Traffic Model for Journal of Electronics and Communication Engineering (IOSR-JECE) GSM/GPRS, Instituto de Telecomunicações/Instituto Superior e-ISSN: 2278-2834,p- ISSN: 2278-8735.Volume 10, Issue 6, Ver. I Técnico, Technical University of Lisbon, 2000. (Nov - Dec .2015), PP 31-35 www.iosrjournals.org [13] K. N. Papaoulakis, D. Nikitopoulos, E. Gkroustiotis, C. Kechagias, C. [3] V. Božidar, Modeling and Characterization of Traffic in a Public Karambalis and G. Karetsos, A Comprehensive Study and Safety Wireless Network, Master of Applied Science Thesis, Simon Performance Evaluation of Operational GSM and GPRS Systems Fraser University, Fall 2006. under Varying Traffic Conditions, Telecommunications Laboratory [4] E.O. Oladeji, E.N. Onwuka and M.A. Aibinu, Determination of Voice National Technical University of Athens Heroon Polytechniou 9, Traffic Busy Hour and Traffic Forecasting in Global System of 15773 Athens, Greece. Mobile Communication (GSM) in Nigeria, 2013 IEEE 11th [14] K. Allen, P. Fitzpatrick and M. Ivanovich, Joint Traffic Signaling Malaysia International Conference on Communications 26th - 28th Capacity Analysis in GSM, Telstra Research Laboratories, Clayton, November 2013, Kuala Lumpur, Malaysia. Vic 3168 [5] S. Kyriazakos, G. Karetsos, E. Gkroustiotis, C. Kechagias and P. [15] M. Boulmalf and S. Akhtar, Performance Evaluation of Fournogerakis, Congestion Study and Resource Management in Operational GSM’s Air-Interface (Um), in Proc. of Applied Cellular Networks of present and Future Generation, IST Mobile Telecommunication Symposium," Orlando, Florida, pages 62-65, Summit 2001, Barcelona, Spain, 9-12 September 2001. March 2003. [6] Z. Jens, Radio Resource Management for Wireless Networks, Artech [16] O.A. Ojesanmi, T.O. Oyebisi, E.O. Oyebode, and O.E. Makinde, House Inc., 2001. Performance Analysis of Congestion Control Scheme for Mobile [7] B. Haider, M. Zafrullah, and M. K. Islam, Radio Frequency Communication Network, International Journal of Computer Optimization and QoS Evaluation in Operational GSM Network, Science and Telecommunications, Volume 2, Issue 8, November Proceedings of the World Congress on Engineering and Computer 2011 Science 2009, Vol I WCECS 2009, October 20-22, 2009, San [17] J. Xiaolong and M. Geyong, Modelling Priority Queueing Systems Francisco, USA. with Multi-Class Self-Similar Network Traffic, Proceedings of [8] I. Kennedy, Lost Call Theory, Lecture Notes, ELEN5007 – ICC'07, pp. 13-19 Teletraffic Engineering, School of Electrical and Information [18] J.E. Flood, Telecommunications Switching, Traffic and Networks, Engineering, University of the Witwatersrand, 2005 New York: Prentice-Hall, 1998. [9] B. Jabari, Teletraffic Aspects of Evolving and Next-Generation [19] Ericsson, Root Cause Analysis for Key Performance Indicators (KPI) Wireless Communication Networks, IEEE Personal Communications for GSM, Networks, REP00271 A, ESA/SK 06:0027. Mag., Vol. 3, No.6, Dec. 1996, pp4-9 249