Maintenance of Voice Quality Control in the Evolution to Packet Switched Wireless Networks N Nageshar, R van Olst School of Electrical and Information Engineering University of the Witwatersrand, Private Bag 3, WITS, 2050 Johannesburg, South Africa email: {nikesh.nageshar@gmail.com; rex.vanolst@wits.ac.za} Abstract- As wireless systems evolve from traditional circuit switch technology to packet based technology there is a prerequisite that voice quality in packet switched networks be maintained to an acceptable level such that user experience is not compromised. All next generation wireless networks have been specified with packet based radio access network (RAN) which implies that the flaws of traditional packet based networks now also apply to wireless networks. This paper illustrates a basic overview of the evolved packet core as well as the quality of service (QoS) mechanisms available on the radio segment of popular packet switched wireless networks. As many carriers have been patient with their take up of IMS, solutions such as circuit switched fallback (CSFB), voice over generic access (VoLGA) and WiMAX voice are presented for consideration. As an example of a packet switched wireless network, a test bed comprising of WiMAX with differing service flows (SF) was configured and the test results for each of the WiMAX QoS service flows presented. The results illustrate superior performance for latency and jitter when utilizing the ertps (enhanced real time Polling Service) as well as the unsolicited grant service (UGS) as compared to the other service flow types. The results for each of the service flow types are compared and contrasted to determine the most appropriate type that is applicable to voice. Index Terms IP, LTE, QoS, VoIP, WiMAX I. INTRODUCTION The development of quality control mechanisms in dealing with voice over a packet switched network is important for the efficient deployment of next generation wireless networks. Within the South African context many local operators have deployed extensive GSM, WCDMA and CDMA2000 networks to cater for voice. The South African situation echoes that of the rest of the continent, whereby majority of the population are mobile telephony users. Among these 40+ million subscribers about 2 million are considered as broadband subscribers with a further 6 million that make use of their mobile telephones for the purposes of e-mail access and basic data [1]. Unlike previous standards the current 3.9 and 4th generation standards have been specified as all-ip networks providing an end-to-end packet based connection for all services [2]. In previous cellular telecommunications standards such as CDMA and GSM voice was inherently the main service offered over dedicated circuit switched channels whereas in the latter standards video, audio and interactive data services are strongly considered [3]. The quality of service provisioning for voice in an IP packed switched network is increasingly difficult because of tight delay, jitter and packet loss demands for voice. This combined with the scarcity of radio resources during bursts of mixed traffic makes the provisioning of voice with a fair to perfect MOS (Mean Opinion Score) a challenge [2]. In respect of packet voice or voice over IP, a network should consist of sufficient bandwidth to carry the coded voice and relevant application, transmission and network protocol overheads. The network should have less than 0.25% packet loss, a maximum jitter of 5 millisecond and less than 150 millisecond packet delay [4]. These parameters have been determined by relating network quality to objective and subjective voice quality metrics. It has been found that a greater than 0.25% packet loss, 5 millisecond jitter and /or 150 millisecond delay significantly contributed to speech stutter, irregular speech and speech delay to a point where the called party can not sufficiently comprehend the calling party. With the slow take up of IMS by operators, if an operator chooses to deploy a fourth generation (4G) or later technology, the operator may be forced to deploy the fourth generation or latter technology as a data only network. Alternatively the operator has the option to deploy a circuit switch fall back mechanism that reverts to the previous generation network in the event that a voice call is made [2] [5]. This however defeats the purpose of convergence and diminishes the cost related advantages of such. The purpose of this paper is to highlight options that will enable quality voice over packet switched wireless networks. The options that shall be illustrated may be challenging in terms of an operators cost to implement such, none the less they are options for consideration. The paper is set out as follows: Section 2 provides an explanation into packet switched wireless networks; Section 3 highlights the current quality of service framework in fourth generation networks. Section 4 is a brief examination of the interim options available to operators for voice over packet switched wireless networks, Section 5 highlights the system model used for testing as well as the results achieved; finally the conclusion is presented in Section 6.
II. PACKET SWITCHED WIRELESS NETWORKS Packet-switched is a description of a network in which small units of data termed packets are routed through the network based on the destination address contained in the header of each packet. Breaking communications down into packets allows the same data channel to be shared among many users in the network. At the entry point of the data channel the conversation is broken down or segmented into packets that are reassembled at exit point of the data channel. Within the wireless realm, fourth-generation (4G) networks consist of a structure that is packet based, termed all-ip. This is where Internet Protocol (IP) packets traverse an access network and a backbone network without any protocol conversion [6]. (MS) and is marked with a connection ID illustrating QoS attributes such as packet latency/jitter and throughput [8]. The IEEE 802.16e supports five SF types [8] [9]. Each of the above SF has various QoS attributes associated. This is indicated in Table I below [9]. Service Flow Type MRTR TABLE I QOS PARAMETERS FOR IEEE802.16E MSTR Max Latency Max Jitter Traffic Priority UGS X X X ertps X X X X X rtps X X X X nrtps X X X BE X X MRTR - minimum reserved traffic rate MSTR - maximum sustained traffic rate Max Latency - maximum packet delay over the air interface Max Jitter - maximum packet variation delay Figure 1: Packet switched wireless network Existing circuit switched cellular networks, consist of base stations (or base transceiver stations), base station controllers, switching centres, gateways, and so on. The base station (BS) does the fast power control and wireless scheduling. The base station controller (BSC) executes the majority of the radio resource management. In contrast, the 4G network has a simple structure where each BS functions intelligently to perform radio resource management as well as physical transmission [6] [7]. A. The Evolved Packet Core Radio access solutions are a primary consideration of the 4G deployment strategy as this will play a central role in service control and the efficient use of network resources. As a result the required network architecture called for a transition to a flat, all-ip core network, called the Evolved Packet Core (EPC). The EPC features a simplified architecture and open interfaces, higher throughput and lower latency. If LTE is considered as an example, the LTE EPC defines a series of new network functions that flattens the architecture by reducing the number of nodes in the network. Figure 1 is an illustration of the LTE EPC [2] [7]. III. QUALITY OF SERVICE (QOS) FRAMEWORK IN FOURTH GENERATION NETWORKS The quality of service (QoS) framework in the standards listed below has been investigated for the purposes of scanning the current QoS specifications and determining its effects on voice quality control: A. IEEE 802.16e (WiMAX) The QoS framework for the IEEE 802.16e standard is based on service flows (SFs). A SF exists between the access service network gateway (ASN-GW) and a mobile station B. LTE The LTE evolved packet system (EPS) is based on a packet flow that is established between the packet data network gateway (PDN-GW) and the user terminal (MS or UE). LTE uses separate service data flows (SDFs) that are mapped to corresponding bearers with a common QoS treatment. LTE offers two types of bearers [8] [9]: Guaranteed Bit Rate (GBR): Dedicated network resources related to a GBR value associated with the bearer are permanently allocated and, Non-Guaranteed Bit Rate (non-gbr): A service utilizing a non-gbr bearer may experience congestion-related packet loss. A SDF within a bearer is assigned a QoS class identifier (QCI). The QCI refers to a set of packet forwarding treatments (e.g., scheduling weights, admission thresholds, queue management thresholds, and link layer protocol configuration) preconfigured for each network element [9]. The QCI characteristics are listed in Table II [9]. The mapping of an SDF to a dedicated bearer is classified by IP five-tuple based packet filter either provisioned in the policy and charging rules function (PCRF) or defined by the application layer signalling [9]. QCI TABLE II QUALITY OF SERVICE CLASS IDENTIFIER (QCI) CHARACTERISTICS ON LTE Resource Type Priority Packet Delay Budget Packet Error Loss Rate 1 GBR 2 100ms 10-2 2 GBR 4 150ms 10-3 3 GBR 3 50ms 10-3 4 GBR 5 300ms 10-6 5 Non-GBR 1 100ms 10-6 6 Non-GBR 6 300ms 10-6 7 Non-GBR 7 100ms 10-3 8 Non-GBR 8 300ms 10-6 9 Non-GBR 9 300ms 10-6
In order enforce QoS for voice the most appropriate characteristic as listed above need to be sufficiently negotiated before a call is made. IMS has been created as an option that can be used to negotiate and support QoS for IP multimedia sessions thereby offering users an enhanced service quality [10]. IMS has been defined as the multimedia session-control subsystem encompassing core network elements for the provision of multimedia services based on a horizontallylayered architecture. The IMS procedures for negotiating multimedia session characteristics are specified by the 3GPP and are based on Internet Engineering Task Force (IETF) Session Initiation Protocol (SIP) [10]. IMS has unfortunately taken longer to realize than originally envisaged by the 3GPP as it has been slow on take up by operators, even after many years of standardisation. B. Voice over LTE Generic Access (VoLGA) Voice over LTE generic access (VoLGA) makes use of a VoLGA Network Access Controller (VANC) to tunnel circuit switched traffic across the LTE network. VoLGA is based on the existing 3GPP Generic Access Standard (GAN) which utilizes virtual packet tunnelling to transport signalling and circuit switched voice to the VANC. The VANC in turn terminates these traffic flows onto the MSC. The limitations of VoLGA are as follows: requires terminal modifications and has not been standardized by the 3GPP [11]. IV. INTERIM OPTIONS FOR VOICE OVER PACKET SWITCHED WIRELESS NETWORKS With respect to the ideal scenario for voice, over packet switched wireless networks, every area from the air interface of the dialled party to the medium of reception of the received party needs to be sufficiently managed to ensure an acceptable QoS. The interim options below does not seek to undermine any current or future technologies that may provide QoS mechanisms for voice but rather provide options to carriers so that they may not be held ransom to such technologies in terms of price or ease of deployment. A. Circuit Switch Fall Back Circuit switch fall back (CSFB) uses the existing 2G/3G network to deliver voice over LTE. The LTE network utilises the paging channel to redirect a voice call from the LTE network to the existing 2G/3G network. The limitations of such a solution are as follows; user terminals need the ability to access both the LTE network and the 2G/3G network, CSFB is signalling intensive hence induces an extended post dial delay [5] [11]. Figure 3: Voice over LTE generic access architecture C. WiMAX Voice As WiMAX consists of multiple QoS service flows over the air interface the best possible way to ensure acceptable voice quality is to utilize the most appropriate QoS SF on the existing WiMAX QoS structure as well as ensure that this QoS is maintained all the way to the core of the network [8] [9]. As an example, voice traffic can be carried from the customer CPE device in a tagged virtual local area network (VLAN) to the WiMAX subscriber station (SS). This voice VLAN can then be forwarded across the air interface to the access service network gateway (ASN GW) with the appropriate WiMAX QoS service flow being applied to the VLAN. Differentiated service code point (DSCP) [12] [13] can then be used between the ASN-GW and the IP switching core, with the voice traffic mapped onto the appropriate DSCP upon exit from the ASN-GW. The limitations are that QoS is manually configured and not automatically triggered; traffic engineering is also required so as to prevent degradation of services that have been marked with a lower QoS SF or DSCP [11]. Figure 2: Circuit switch fallback architecture
UGS service flow performed the best with the BE service flow performing the worst. Although all of the service flow types performed well below the required latency thresholds, it has to be noted that such latency figures are considered as significantly high if compared to a fibre link along the same path. This may adversely affect inter-continental voice calls when combined with the latency associated to satellite or cable links. 40 35 30 Latency (ms) 25 20 15 Figure 4: Voice over WiMAX test model 10 5 0 UGS ertps rtps nrtps BE WiMAX QoS Parameters - Service Flow Type V. SYSTEM MODEL AND RESULTS It would have been ideal to test each of the illustrated options above, however WiMAX was selected due to its ease of availability at the time of conducting the experiments. The WiMAX test bed presented consisted of the following components: 1. A WiMAX subscriber station (SS) with access to the WiMAX network management element. 2. Access to a WiMAX base station. 3. Connectivity to an IP network. 4. CPE devices on either side of the WiMAX link. 5. PC s at either side of the WiMAX link. The system model was configured in the following manner; the first PC was connected to the WiMAX SS. The WiMAX link was set up with a 2 Meg profile between the WiMAX SS and the WiMAX base station. The WiMAX base station was located about 2-3 Km away from the WiMAX SS. The WiMAX base station was thereafter connected via a local IP LAN to the second PC in the lab. For the purposes of this experiment; 2 Meg profiles on the uplink were created each with of the following service flows types: unsolicited grant service (UGS), enhanced real time polling service (ertps), real time polling service (rtps), nonreal time polling service (nrtps) or best effort (BE). Upon enforcement of the relevant service flow type, the configured link was flooded with both UDP and TCP packets based on a constant 2 Meg profile. Latency, jitter and packet loss were measured using the IPERF network testing tool [14] and multiple ping tests. The results of the IPERF network testing tool and ping tests are listed in Figure 5 and 6 respectively. An estimated 10 independent tests were run for each of the service flow types. Each test comprised of 20 send receive responses. A. WiMAX Latency results The latency results for each of the profiles ranged between 28ms and 37ms. This was well below the 150 ms one way or 250ms round trip time thresholds for voice. The Figure 5: Latency on differing WiMAX service flow types B. WiMAX Jitter results The Jitter results obtained for each of the profiles ranged between 3.5ms and 5.4ms with the ertps service flow performed the best and the rtps service flow performing the worst. It was also noted that the performance of the rtps, nrtps and BE service flows were similarly grouped confirming that jitter was not a prioritised for these service flows as depicted in Table 1. Jitter (ms) 6 5 4 3 2 1 0 UGS ertps rtps nrtps BE WiMAX QoS Parameters - Service Flow Type Figure 6: Jitter on differing WiMAX service flow types The results for packet loss were not illustrated for the reason that across all the service flows packet loss was virtually zero with the occasional result of a less than 0.03% packet loss. When considering the most appropriate service flow for voice: rtps, nrtps and BE either fell below or were too close to the 5ms jitter threshold for voice, hence these service flows were immediately ruled out. The UGS and ertps service flows were deemed the most appropriate for voice however the ertps service flow 5 ms
became the preferred option due to it demonstrating a 20% better jitter performance that the UGS service flow. VI. CONCLUSION The ability to successfully provide quality voice on next generation packet switched wireless networks is vital. Vendors and standards bodies have taken liberty to provide efficient data networks consisting of high throughputs with converged services being the central task as opposed to voice. This is turn has left operators hunting for possibilities to provide voice over their newly acquired packet switched wireless networks. The purpose of this paper is to highlight the possibilities that operators currently have to successfully carry quality voice over next generation packet switched wireless networks. The various QoS mechanisms on 4G standards have been briefly examined illustrating the possibility of carrying voice within the QoS frame work. IMS Circuit switch fallback, VOLGA and WiMAX voice were also discussed and presented as interim options. WiMAX service flows (SF) with differing QoS parameters were tested with the test results for each of the QoS parameters illustrating greater performance in terms of latency and jitter when utilising the ertps (enhanced real time Polling Service) and the unsolicited grant service (UGS) as compared to the other service flow types. The varying quality of service class identifiers (QCI) within LTE service data flows are planned to be tested in a similar fashion. Finally it can be said that voice can be successfully carried over a packet switched wireless network provided that the correct architecture or interim option in chosen and the most appropriate QoS attributes are in place. ACKNOWLEDGMENT This work was supported in part by the University of Witwatersrand School of Electrical and Information Engineering. The authors acknowledge the support of the University of the Witwatersrand and Neotel South Africa. REFERENCES [1] Goldstuck A. World Wide Worx, http://www.worldwideworx.com/, accessed 1 June 2010. [2] 3GPP TS 36.300 V9.3.0. Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Network (E-UTRAN); Stage 2, March 2010. [3] Rein S, Fitzek F, Reisslein M. Voice Quality Evaluation in Wireless Packet Communication Systems: A Tutorial and Performance Results for ROHC, IEEE Wireless Communications, February 2005, pp. 60-67. [4] Schutte J, Helberg A. A Study of the Effect of MPLS on Quality of Service in Wireless LANS, South African Institute of Electrical Engineers Journal, Volume 99, September 2008, pp. 70-76. [5] 3GPP TS 23.272 V9.3.0. Circuit Switched (CS) Fallback in Evolved Packet System (EPS); Stage 2, March 2010. [6] Choi Y, Lee K, Bahk S. All-IP 4G Network architecture for efficient mobility and resource management, IEEE Wireless Communications Journal, May 2007, pp 42-46. [7] 3GPP TS 36.201 V9.1.0. Evolved Universal Terrestrial Radio Access (E-UTRA); LTE physical layer; General description (Release 9), March 2010. [8] IEEE 802.16 2009. Part 16: Air Interface for Broadband Wireless Access Systems, May 2009 [9] Alasti M, Neekzad B, Hui J, Vannithamby R. Quality of Service in WiMAX and LTE Networks, IEEE Communications Magazine, May 2010, pp. 104-111 [10] Skorin-Kapov L, Mosmondor M, Dobrijevic O, Matijasevic, M. Application-Level QoS Negotiation and Signalling for Advanced Multimedia Services in the IMS, IEEE Communications Magazine, July 2007, pp 108-116. [11] Unstrung Insider. Voice over LTE: Many Questions, No Easy Answers, Light Reading, Volume 8, No. 10, October 2009. [12] IETF RFC 2474. Definition of the Differentiated Services Field (DS Field) in the IPv4 and IPv6 Headers, December 1998. [13] IETF RFC 2475. An Architecture for Differentiated Services, December 1998. [14] NLANR/DAST, IPERF testing Tool, Sourceforge http://sourceforge.net/projects/iperf/?abmode=1 Nikesh Nageshar, is currently a student at the University of the Witwatersrand, Johannesburg, South Africa and is employed at Neotel, South Africa as an engineer at the Technology and Innovation division (e-mail: nikesh.nageshar@gmail.com). Rex Van Olst, is currently an Associate Professor with the University of the Witwatersrand, Johannesburg, South Africa. He heads the Telecommunication postgraduate research arm at the University of the Witwatersrand School of Electrical and Information Engineering (e-mail: rex.vanolst@wits.ac.za).