Providing Voice Connectivity to Rural India using WiMAX: Issues and Solution

Transcription

1 Providing Voice Connectivity to Rural India using WiMAX: Issues and Solution Abhijit Lele Motorola India Research Labs Bangalore, INDIA Mayank Raj International Institute of Information Technology, Bangalore, INDIA Debabrata Das International Institute of Information Technology Bangalore, INDIA ABSTRACT The need of emerging markets telecom, specifically in the Indian telecom scenario, WiMAX is being looked as a broadband access solution ahead of LTE and other competing technologies due its long range and high bandwidth. Voice Over IP (VoIP) will potentially be the killer application for emerging market like India. In this paper we propose a Kiosk based WiMAX infrastructure model to provide voice connectivity to rural Indian villages. In the proposed kiosk model, plain old telephones are connected to a WiMAX subscriber station using Foreign Exchange Subscriber and a Media Gateway. The novelty of the kiosk based infrastructure models is that it has low deployment cost from a service provider perspective, and almost negligible equipment cost for the end user. In order to make the kiosk based model economically sustainable, the number of simultaneous voice calls that can be supported over the WiMAX subscriber stations needs to be maximized. To this end, the paper proposes a Dynamic Frame Profile algorithm to maximize the number of VoIP calls supported over a single subscriber station. A performance evaluation of the the proposed Dynamic Frame Profile algorithm is also carried out to study its effectiveness and reported in this paper. 1. INTRODUCTION Connectivity, in general is vital to any business and society. More so for emerging markets like India. Although only half a percent of Indian population has residential Internet access, about million[1] telephone subscribers exist in India as of March 26, which translates to approximately 1 telephone line per 1 users. Out of the million subscribers, about million are fixed wireline telephone subscribers, and 1 million are mobile subscribers. Fixed telephone subscriber growth has had only marginal growth as compared to mobile subscribers due to the following two reasons. First, the cost of deployment is of fixed mobile telephones is about USD $6 as compared to USD $1[2] for Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. NSDR 7, August 27, 27, Kyoto, Japan. Copyright 27 ACM /7/8...$5.. mobile phones, making it difficult for telecom operators to break even in terms of revenue. Secondly, the geographical and terrain conditions in rural areas in India is not conducive to laying optical and copper cable, thus making is fixed wireline deployment even more expensive. In order to close the gap of providing voice connectivity between the developed and developing nations, the Government of India came up with a scheme of Public Call Office (PCO). PCOs are small kiosks where a user can go and access telephone services. The PCOs/kiosk model also provide a cost effective way to provide voice connectivity especially in rural areas. Considering the geographical and economic constraints of providing voice connectivity to rural areas in India, use of wireless networks such as in mesh mode[3] are being investigated to provide voice connectivity. In this paper, we propose the use of WiMAX TM 1 based network to extend voice connectivity to rural areas. Learning from the success of the PCO model, we extend this model in this paper, and propose a Kiosk based WiMAX infrastructure model for rural voice connectivity. The rest of the paper is organized as follows. In Section 2 we motivate the need for the use of WiMAX to provide voice connectivity in rural areas in India. The Kiosk based infrastructure model which uses media gateway to provide voice connectivity is discussed in Section 3. The technical challenges involved in Kiosk based model are discussed in Section 4. In order to overcome some of the technical challenges, an enhancement to WiMAX, called as Dynamic Frame Profile algorithm to increase VoIP capacity is discussed in Section 5. A performance evaluation of Dynamic Frame Profile algorithm is reported in Section 6. We finally conclude in Section BACKGROUND An interesting perspective to understand the telecom scenario in India, is the categorization of Indian telecom services. For the purpose of licensing and managing telecom services, the geographical area of India is divided into four tiers [1]. Metros: The large cities in India. These comprise of 4 cities. Tier 1: Major Cities in India. Tier 2: Sub-urban areas in India. Tier 3: Villages in India. 1 WiMAX is a Registered Trademark of WiMAX Forum

2 Table 1 gives the tier wise split of the population in India [2]. The majority of revenues and expected growth rates are now Table 1: Tier Wise Population of India Type of Region Population (in thousand) Metro 52,64 Tier 1 366,84 Tier 2 513,199 Tier 3 219,385 infrastructure must be compatible with existing Plain Old Telephone System (POTS). For this purpose, we propose the use of a V5.2 Media Gateway. The following sections describes in detail the planned infrastructure for providing voice connectivity. 3. RURAL KIOSK MODEL A typical rural village scenario is given in Figure 1. The coming from traditionally under-served Tier-2 and Tier-3 regions, and hence the focus on rural connectivity has never been more intense. Some of the characteristics of Tier-2 and Tier-3 regions are as follows. Low Literary: The average literacy rate is Tier-2 and Tier-3 regions is low, hence providing broadband Internet access will have limited revenue earning. The major revenue earnings for operators will come from voice communications or providing telephone connectivity. Low Income: As per the economic census report the per capita income in Tier-2 region is about USD $245, and in Tier-3 region is USD $19, thus providing low cost telecom services, especially voice is a challenge. PC Penetration: Though the PC penetration in Tier-1 and Tier-2 regions is on the rise, still it is low due to low literacy and income. This again reiterates the fact that providing voice connectivity will be the major focus for all telecom operators. Figure 1: A Typical Rural Scenario rural scenario comprises of a few hundred house holds, or for larger villages in India, a few thousand house holds. As per the discussion in Section 2, the aim is to provide POTS connectivity at an affordable cost to the end user as well as provide a cost effective deployment for the service provide. For this purpose, we propose the use of a kiosk model as shown in Figure 2. In the kiosk based model, each kiosk is a Based on the above discussion it is clear that, in addition to broadband access, voice communications is the primary driver for telecom operators in the rural areas. However, providing this rural voice connectivity has its own set of challenges. As per the TRAI report [1], an ambitious plan to add an additional 7.5 million subscribers over the next 2 months in Tier-2 and Tier-3 regions, which translates to about 379, subscribers every month. Building a fixed wireline infrastructure for this is not an easy task and not economical. The terrain conditions are not suitable for easy deployment of optical fibre or copper cables. Hence there is a need to look at use of wireless networks as backhaul networks. 2.1 In Summary Based on the discussion in Section 2, it is clear that enabling voice communication takes precedence over other services. Considering the terrain conditions it is our belief that wireless networks is the right options to provide voice connectivity to rural areas, especially Indian villages. The other challenge is to reduce the deployment cost. One of the possible approaches to reduce the deployment cost is to use the kiosk model which we propose in this paper. Also, given the fact that WiMAX is being used as an access technology, using WiMAX Customer Premise Equipment (CPE) is not an option for rural environment due to its high cost. Hence, the Figure 2: The Proposed Kiosk Model For Rural Voice Connectivity WiMAX Subscriber Station (SS), and provides voice service to set of houses (POTS) as shown in Figure 2. Each village will be served by a single WiMAX Base Station (BS) or alternately each WiMAX BS sector will service a village. To enable the POTS to be connected to the SS at the kiosk we, propose the use of Foreign Exchange Subscriber (FXS) [4] interface and a V5.2 Media Gateway [5] as shown in Figure 3. The FXS is placed at the Kiosk along with the SS. All the POTS serviced by the SS are terminated at the FXS. The FXS is responsible to convert the analog telephone call to a VoIP call. The Media Gateway which is based on the Media Gateway Control Protocol (MGCP) is responsible to provide the signalling support and interface with the PSTN

3 and FXS. The cost of the SS is approximately USD $8 and a 24 port FXS costs approximately USD $12. Considering additional wiring cost, the total investment is approximately USD $3. The WiMAX BS cost is approximately USD $3, which is assumed to be provided by the primary service provider. The end user has to invest in just purchasing a POTS instrument which costs about USD $5, thus making it a lucrative proposition for the end user. Figure 3: The Proposed Rural Telecommunication Architecture network. In the proposed architecture is designed for fixed line connectivity, however in the future the architecture can be extended to support mobile connections. A typical abstract view of the call flow for the proposed architecture is given in Figure 4. From Figure 4 observe that the end user Figure 4: A Typical Call flow Diagram for the Proposed Architecture (User 1) is trying to make a PSTN call to User 2. The FXS is responsible to convert the analog or DTMF signalling to an VoIP message, which in this case is a Session Initiation Protocol (SIP) invite. This SIP invite is received by the SS, and forwarded to the BS, which in turn forwards to the receiving SS over a WiMAX frame. An Enum server is used at the SS to convert the PSTN number to an equivalent SIP address. Once the receiver (User 2) receives the ring, the loop back signal (Answer) is transmitted back as a SIP OK message to the caller (User 1). Detailed messaging is not discussed as a part of this paper, as it is out of scope of this paper. 3.1 Discussion It is our belief that the above proposed model is not only going to be cost effective and practical for deployment for both the service provider and the end user for the following reasons. The service provider has to invest in the Kiosk, viz SS Since, the end user does not invest much in equipment, it is relatively easy for the service provider to make an affordable business model to recover the investment cost, and a break even period of approximately 2 years is a reasonable estimate. 4. TECHNICAL CHALLENGES Having discussed in Section 3 one of the possible architectures for providing voice connectivity in rural India using WiMAX, this section focusses on the technical challenges involved in deploying the proposed infrastructure. WiMAX, which is based on 82.16[6] air interface standards. In the present architecture, we assume the use of 82.16e[7] standard and 82.16e supports five classes of traffic, out of which two, viz. Unsolicited Grant Service (UGS) and Extended Real-Time Polling Service (ertps) are used for VoIP traffic. From a service provider perspective, the challenge is to maximize the simultaneous number of users supported by the BS. Since the communication between two subscriber stations is VoIP based, the problem statement reduces to maximizing the VoIP capacity of the WiMAX network by enhancing the simultaneous VoIP calls supported by the SS. Our experiments indicate that a maximum of 1-12 simultaneous VoIP calls can be supported over a single subscriber station [8] without degradation in Quality of Service (QoS) 2. In order to maximize the the number of VoIP calls we propose a Dynamic Frame Profile (DFP) algorithm. The algorithm can be implemented without having any extra overhead of messaging and can be incorporated in the current MAC architecture as recommended by standard [6]- [7] without any modifications. The PHY Synchronization Field in DL- MAP sub-frame is used by the BS to inform the SS about the current frame length. The PHY Synchronization Field has a frame length field defined with codes from x to xff which represnt the frame lengths respectively in msec. The first 3 fields are reserved for frame lengths of.5msec, 1msec and 2msec. The rest of the fields are left for the service provider to define, and we propose the use of these fields to convey the frame duration. For the sake of completeness we briefly describe some of the related work in the following section. 4.1 Related Work A good technical overview of 82.16x based systems is given in [9]. A generic performance evaluation of is discussed in [1]. An analysis and discussion on the various scheduling algorithms used in 82.16e standards is discussed 2 For the purpose of brevity, we omit details related to these VoIP performance results

4 in [11]. Two new scheduling policies are analyzed and performance evaluation reported in [12]. The performance of VoIP traffic on an test bed from Alvarion [13] is reported in [14]. In the following section we discuss the proposed DFP algorithm. 5. DFP ALGORITHM The basis of this algorithm is to dynamically change the frame duration based on VoIP load per SS. Let T f denote the default frame duration. Let T s denote the OFDM symbol duration, and hence the total number of symbols S t per frame is given by Equation 1. S t = T f T s (1) Let b s denote the bandwidth allocated per symbol, and hence the total bandwidth B F per frame is given by Equation 2. B f = b s S t (2) Let α ugs and denote α ert denote the fraction of the bandwidth reserved for UGS and ertps class of traffic, then the total bandwidth reserved for VoIP B voip traffic is given by Equation 3. B voip = α ugs B f + α ert B f (3) If B i denotes the bandwidth requested by i th VoIP session, the average bandwidth for VoIP session is given by Equation 4, where k denotes the total number of active VoIP sessions. i=k i=1 B avg = Bi (4) k The average number of N voip VoIP sessions supported by a single SS is given by Equation 5. N voip = Bvoip B avg (5) For an emerging market scenario, N voip needs to be maximized. The DFP algorithm proposed in this paper monitors the VoIP call blocking probability P b. Note that the calls are blocked when the requested bandwidth B i for a service flow cannot be allocated, based on the call admission policy. From Equation 1 and Equation 4, it is clear that N voip is maximized by increasing S t, which implies changing the frame duration T f. Let β denote the frame scaling factor, and B j rej denote the bandwidth requested by the jth VoIP session that has not been admitted into the network. The proposed DFP algorithm is given in Table 2, where B ins and T i denote the instantaneous bandwidth and time respectively. The call drop probability is calculated over periodic time intervals of T h. Instantaneous call drop probability does not give an accurate reflection of the current state of the network and decisions based on the instantaneous call drop probability lead to frequent changes in frame length, leading to degradation in QOS as the DFP algorithm does not converge. In DFP algorithm, the call drop probability is calculated over a window of time. Let P prev denote the call drop probability at the t δ time instant and P b denote the over all call drop probability at time instant t. While calculating P b, appropriate weight factors are used. Steps 3-9 in Table 2 compute the number of accepted and rejected calls in the current time interval and the total bandwidth required for the rejected sessions. On admission of a call if Table 2: Pseude Code of DFP Algorithm 1. Set Frame Length = T f, B rej =, P prev = 2. if (T i T h ) 3. On Arrival i th VoIP session 4. If (B i + B ins B voip ) 5. Admit Session i 6. Update call admission Count 7. if l i l th 8. Recompute frame length 9. Endif 6. else 7. Reject Session 8. Update Call drop count 9. B rej = B rej + B i 9. Endelse 1. Endif 11. else 12. Compute instantaneous P i 13. if (P i = ), set P i = P prev 15. Endif 16. P b = W 1 P prev + w 2 P i, where < w 1 w 2 < P prev = P b 18. If P b P th 19. Set T new f 2. Set T f = Tf new 21. Endif 22. Endelse 23. Endif = T f + β [ B rej T s b s ] the load per SS exceeds a threshold load per SS l th, a new frame duration is calculated. On completion of the current time interval the instantaneous call drop probability is calculated. If no calls arrive in the current time duration or no calls are dropped the instantaneous call drop probability P i is zero. In such a case the instantaneous call drop probability is re-initialized to the previous call drop probability P prev. This give the algorithm to maintain the current state of frame duration for a longer time interval, thus stablizing it. Step 16 in Table 2 calculates the weighted average call drop probability with weights < w 1 w 2 < 1. Once the call blocking probability exceeds the threshold call blocking probability P th, a new frame duration Tf new is computed in step 19 using Equations 1 and Equation 2. The following section reports the performance evaluation of the DFP algorithm. 6. PERFORMANCE OF DFP ALGORITHM A performance evaluation using Network Simulator (NS- 2) [15] patched with the WiMAX extension proposed in [16] is carried out to study the effectiveness of the proposed DFP algorithm. Simulations are carried out for three different scenarios. Scenario 1: Voice carried only over UGS. Scenario 2: Voice carried only over ertps. Scenario 3: Voice carried over UGS + ertps. The simulation scenario comprises of a single BS and 8 SS. The call duration is poisson distributed with an average of

5 vs. Call Blocking Probability (ertps) vs. Call Blocking Probability (UGS + ertps) Call Blocking Probability Call Blocking Probability (a) UGS (b) ertps II (c) UGS + ertps Figure 5: Call Blocking Probability for UGS, ertps, and UGS + ertps vs. Avg. Delay Jitter for UGS vs Agv. Delay Jitter (ertps) vs Avg. Delay Jitter (UGS + ertps) Avg. Delay Jitter (msec) Avg Delay Jitter (msec) Avg. Delay Jitter (a) UGS (b) ertps II (c) UGS + ertps Figure 6: Delay Jitter for UGS, ertps, and UGS + ertps vs. Packet Drop Probability for UGS vs. Packet Drop Probability (ertps) BS Buuffer Size vs. Packet Drop Probability (UGS + ertps) 9.E Packet Drop Probability 8.E-2 7.E-2 6.E-2 5.E-2 4.E-2 3.E-2 2.E-2 1.E-2 Packet Drop Probability Packet Drop Probability E (a) UGS (b) ertps II (c) UGS + ertps Figure 7: Packet Loss Rate for UGS, ertps, and UGS + ertps

6 16 seconds. The inter arrival time between calls is exponentially distributed with an average time of.1 seconds. Such high loads of approximately 64 connections were considered to get realistic simulations results for a rural scenario. The results are averaged over multiple runs of different frame durations. Figure 5 gives the call blocking probability for the three scenarios for different buffer size (in terms of packets) at the BS. The following conclusions can be drawn: From Figures 5 observe that for a BS buffer size of 25, the call blocking probability tends to zero, indicating that by varying the frame duration, the number of simultaneous VoIP calls that can be accommodated increases. Another important metric for VoIP based applications is the delay jitter and packet loss probability. When, the number of VoIP calls increase, the delay jitter should be at most 5 msec and the packet loss probability should be within negotiated QoS bounds. Figure 6 gives the delay jitter for the three scenarios. From Figure 6(b) observe that with increase in buffer length, the delay jitter increases. Note that in all the three scenarios, the delay jitter is well within the tolerable limits of 5 msec. Hence, varying the frame length does not impact the quality of the VoIP calls. Figure 7 plots the packet loss probability for the three scenarios. The following conclusions can be drawn from Figure 7. From Figure 7(a), observe that the packet loss rate increases with increase in frame duration and reduces with increase in buffer size at the BS. In general from Figure 7, observe that the packet loss rate increases with increase in frame duration and reduces with increase in buffer size at the BS. Note that, in the present embodiment as the frame duration increase beyond 8 msec, the packet loss rate increases beyond the threshold value. Hence, for the present scenario the frame duration cannot be increased beyond 8 msec. 6.1 Discussion The key to the kiosk based model is to maximize the number of VoIP calls per SS. The above given results give us guidelines to configuring the SS and BS for this purpose. From Figure 7, an optimal BS buffer size can be selected to reduce Call blocking probability and hence effectively increase the number of VoIP calls. For this selected BS buffer size, the exptected delay jitter per packet per active connection can be obtained from 6. Thus, in general these performance figures gives the guidelines to obtain an optimal parameters. 7. CONCLUSION In this paper we have proposed a Kiosk based infrastructure model to provide voice connectivity in rural India. The salient features of the Kiosk based infrastructure model include, relatively low deployment cost and economically viable for the end user. The key to the success of the proposed Kiosk model is to maximize the number of voice calls supported over a single WiMAX subscriber station. To this end, the paper proposes and evaluates a Dynamic Frame Profile algorithm. It is our belief that using the Kiosk model, it will be feasible to provide voice connectivity to rural India in a cost effective way. 8. REFERENCES [1] Telecom Regulatory Authority of India. 26 Annual Report from Telecom Regulatory Authority of India. India, 26. [2] Sidearm Pad and Alkane Delles. India Broadband Wireless and WIMP Market Analysis and Forecasts: Maravadis Telecom MarketResearch and Analysis, 26. [3] D.P. Hole and F.N. Tobagi. Capacity of an ieee 82.11b wireless lan supporting voip. In Proc.of the IEEE International Conference on Communications, pages Vol. 1. IEEE, June 24. [4] Introduction to VoIP Technology and Applications. [5] RFC 366. Basic Media Gateway Protocol. [6] air interface standard specification. [7] IEEE 82.16e 24. In IEEE Standard for Local and Metropolitan Area and System Mobility Support, February 26. [8] Mayank Raj et al. Performance evaluation of voip over 82.16e - an emerging market perspective. Submitted to Globecomm, November 27. [9] WiMAX Forum. Mobile wimax - a technical overview and performance evaluation. In Technical Report from WiMAX Forum, August 26. [1] Fan Wang et al. Ieee 82.16e system performance: Analysis and simulations. In Proc.of the IEEE 16th International Symposium on Personal, Indoor and Mobile Radio Communications, pages 9 94 Vol. 2. IEEE, September 25. [11] Howon Lee et. al. Performance analysis of scheduling algorithms for voip services in ieee 82.16e systems. In Proc. of the IEEE Vehicular Technology Conference, pages Vol. 3. IEEE, May 26. [12] S. Hong et al. Considerations of voip services in ieee broadband wireless access systems. In Proc. of the IEEE Vehicular Technology Conference, pages Vol. 3. IEEE, May [13] Alvarion Technologies. [14] Nicola Scalabrino et al. Performance evaluation of a wimax testbed under voip traffic. In Proc. of ACM Conference on Wireless Networks, pages ACM SIGMOBILE, September 26. [15] The Network Simulator. [16] Jenhui Chen et al. The design and implementation of wimax module for ns-2 simulator. In Proceedings of NS2 Conference, October 26.