VIEWS: 25 PAGES: 6 POSTED ON: 12/12/2010
VoIP Service on HSDPA in Mixed Trafﬁc Scenarios Yong-Seok Kim Media Lab., Telecommunication R&D center Telecommunication Network Business Samsung Electronics email@example.com Abstract IP Network UE NodeB RNC SGSN/GGSN In this paper, we evaluate the capacity of voice over UL delay : 40ms UL Processing delay : 30ms Backhaul delay : 30ms IP Network delay : about 42ms internet protocol (VoIP) services on high-speed downlink NodeB-GSN:15ms, GSN-NodeB:15ms packet access (HSDPA), in which frame-bundling is incor- DL Processing delay : 30ms DL delay : Variable porated to reduce the effect of relatively large headers in the IP/UDP/RTP layers. Also, the capacity of VoIP service on HSDPA is considered in a mixed voice and best-effort Figure 1. End-to-end delay component trafﬁc scenarios. In order to guarantee QoS for VoIP ser- vice, a design of packet scheduler based on the proportional DPA . This is due to some inherent spectral inefﬁcien- fairness scheme is provided. Simulation results show that cies that seemingly would disappear if both voice and data performance of VoIP service with frame-bundling scheme is were carried on 3G wireless radio link. Backhaul facilities highly sensitive to delay budget. We also conclude that the would be more efﬁcient because voice, data trafﬁc and all capacity of VoIP service on HSDPA is attractive for trans- signaling protocols would be carried on same IP facilities. mission of voice, if compared to the circuit switched voice Hence, in this paper, it is attempted to provide system-level and to the VoIP on WCDMA used in Release’99. simulation results of VoIP capacity which reﬂects the frame bundling (FB) and a best-effort (BE) service. This paper is organized as follows. In Section II, we present the system 1. Introduction of VoIP on HSDPA. And the simulation conﬁguration is de- scribed in section III. Simulation results and conclusion are High-speed downlink packet access (HSDPA) system presented in Section IV and Section V respectively. that supports the peak rate of 14.4 Mb/s outperforms the third generation (3G) WCDMA system speciﬁed in 3GPP Release’99 . The transmission of voice using packet 2. VoIP Service on HSDPA data internet protocols (IPs) is arguably the hottest atten- tion in telecommunication technology today. This is be- 2.1. Traﬃc model and protocols cause it has high visibility in the consumer space. A long- distance VoIP calling is cheaper to operate, maintain and In this work, two trafﬁc models are considered for the upgrade than comparable solutions using switched digital cases of BE service and VoIP service. In the context of or analogue phone service. In addition, it facilitates the cre- BE trafﬁc, we applied the full queue trafﬁc model assuming ation of new services that combine voice communication that data can be always sent when a queue of certain user with other media and data applications such as video and is chosen. A telephone conversation can be represented by ﬁle sharing . Early VoIP studies were focused on the ON/OFF patterns. ON periods correspond to a conversant wireless local area networks (WLAN) because of its con- talking and OFF periods are due to silences. The duration venience and achievable high-speed data rate as that of the of both ON and OFF periods is negative exponentially dis- wireline network . Despite the success of VoIP in wire- tributed with an average of appropriate seconds. Generally, line and WLAN networks, the most widely held expectation conversation trafﬁc can be approximated to the two state is about introducing into the latest broadband 3G technolo- Markov trafﬁc model with a suitable voice activity factor gies such as CDMA2000 1X EV-DO  and WCDMA HS- . The adaptive multirate (AMR) voice codec is manda- Table 1. Summary of end-to-end delay com- Table 2. Simulation parameters (HSDPA) ponent Parameter Assumption Source trafﬁc AMR 12.2kbps, Delay component Delay assumption packet overheads voice activity=0.32, 2-state Markov Voice encoder 20ms(12.2Kbps) Bundling(FB0-FB2) RTP frame bundling FBx(20×xms) ROHC 3 bytes [IETF RFC 3059] NodeB scheduling+HARQ Correspond to FB Cellular layout Hexagonal grid, 19 sites, 3 sectors (Max.110ms if FB0) (NB-to-NB 1km) NodeB 30ms Carrier frequency 1.9GHz ROHC, RLC+MAC processing Propagation loss Path loss=-128.1-37.6*log(R) Downlink propagation Shadowing model Log Normal Std. dev. 8dB UE scheduling+HARQ 40ms [Hata model] UE 30ms UE speed 3km/h(50%)+120km/h(50%) processing,buffering,etc Antenna gain Node B 14dB / UE 0dB Uplink propagation Other loss -10dB Backhaul delay (GSN-NodeB) 30ms Fading Model Combination Ped.A(5%) IP network delay About 42ms +Ped.B(45%)+Veh.B(50%) Evaluated with 3GPP (TS 25.101) UE Rx diversity with or without considered (catagory 10) tory for voice services in WCDMA systems. During bursts Retransmission No RLC retransmission of conversation, with the AMR mode of 12.2kbps, the VoIP HARQ (max. retrial = 6) application generates 32-bytes voice payload at 20ms inter- CQI delay, error 3TTI (6ms), 1% vals . During silent periods, a 7-bytes payload carries a Scheduling Modiﬁed PF scheduler silence descriptor (SID) frame at 160ms intervals. A typi- Reserved - Common channel cal VoIP protocol stack, which employs the real-time trans- power overhead 20% port protocol (RTP), is encapsulated to the user datagram - Associated DCH power protocol (UDP). This, in turn, is carried by IP. The com- 0.3% per mobile user bined these protocols demand a 40-bytes IPv4 header or a - HS-SCCH power 9dB offset 60-bytes IPv6 header. Obviously, the overhead caused from via associated DCH the header to support VoIP service seriously degrades the - common channel code 10 spectral efﬁciency. Therefore, efﬁcient and robust header - Associated DCH code 1 compression (ROHC) technique can be used to reduce the per mobile user effect of relatively large headers in the IP/UDP/RTP lay- - HS-SCCH codes are considered ers. This technique can reduce the size of the IP/UDP/RTP headers as little as 2 or 4 bytes. Maximum compression 1 byte can be achieved by imposing limitations . more packetization delay is introduced but at the same time 2.2. Frame bundling overhead is more and more reduced. For conversational ser- vices low delay is crucial, hence only few payloads are bun- In contrast to BE trafﬁc, voice packets are usually very dled. So, in this paper, the employed maximum number of small. Hence, packing a single payload to a RTP packet data FB is 2 (typically up to 2 in the case of AMR), represented unit (PDU) may introduce severe overhead. For example, by FB2. Where, FBx means that the number of adopted FB compared to 40 bytes IPv4 header 32 bytes payload would is x. These FB features are supported in the AMR payload be a typical case for codec used for AMR. To decrease the format . overhead, instead of adding a single payload multiple con- secutive payloads are packed into a single RTP PDU. Also, 2.3. End-to-end delay budget for QoS sup- FB can be used to decrease the occurrence of bit-stufﬁng port due to the mismatch between the size of VoIP packet and that of HSDPA media access control (MAC) frame format To ensure end-to-end QoS in a packet-switched (PS) net- . However, there is a trade-off between delay and over- work, the low delay is one of the most important criteria head. That is, the more packets are bundled together the for maintaining high-quality VoIP service. But, to attain 1.0 1.0 0.95 0.95 0.9 0.9 Percentage of UEs with BLER<2% Percentage of UEs with BLER<2% 0.85 0.85 0.8 0.8 0.75 0.75 0.7 w/ UE Rx Diversity 0.7 w/o FB w/ FB1 0.65 0.65 w/ FB2 w/o UE Rx Diversity 60UEs 0.6 0.6 70UEs 80UEs 0.55 0.55 60UEs w/o UE Rx Diversity 100UEs 110UEs 100UEs w/ UE Rx Diversity 0.5 0.5 10 20 30 40 50 60 70 80 90 100 110 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 Scheduler Delay Budget [msec] Scheduler Delay Budget [msec] Figure 2. Outage versus Delay budget for dif- Figure 3. Outage versus Delay budget for dif- ferent UEs w/ or w/o UE receiver diversity ferent value of FB high VoIP capacity, the scheduler must have sufﬁcient time (TTI). The use of effective scheduling algorithm is neces- to manage voice packets. Of the assumed 285ms end-to- sary for improving the throughput of system, since the HS- end delay budget for qualiﬁed voice service, about 110ms DPA system shares resources with multi-users at the same is available for scheduling in the downlink . The delay TTI. The HSDPA system was designed for services with in IP and backhaul network is in general bounded to 72ms high throughput for BE trafﬁc requirements but it can also [ref] which is ﬁxed value allowing us to focus on the delay be used for VoIP. Generally, VoIP regarding conversational budget within radio access network. The end-to-end delay service has the highest priority of trafﬁc classes as it is very budget in the case of mobile-to-mobile conversation can be sensitive to the transmission delay. The use of separate pri- assumed as Fig. 1. Although VoIP performance depends on ority queues makes it possible to optimize HSDPA schedul- both downlink and uplink performance, we would like to set ing for each different QoS service. This paper optimizes the aside the consideration of both directions as comprehensive scheduling algorithm for trafﬁcs demanding different QoS study for future research. Table 1 summarizes the assump- by using priority handling. First of all, VoIP ﬂows must tion of each component in terms of end-to-end delay budget be handled by higher priority than interactive trafﬁc. Node . B receives the value of priority from the radio network con- troller (RNC). Hence, we can differentiate between services 2.4. Performance criteria for VoIP through the priority values. Meanwhile, for VoIP service over HSDPA, it is beneﬁcial to include time-delay factor in The main objective of this paper is about the VoIP ca- scheduling algorithm. To measure delay, the scheduler puts pacity in the sense that there exists the maximum number time-stamp in each packet as it arrives at the priority queue of VoIP users that can be supported per sector without ex- that is operated with 8 queue buffers to support different ser- ceeding a given outage threshold. In PS network, packets vice. Therefore, in this paper, a modiﬁed proportional fair will be dropped under network trafﬁc loads congestion due (PF) scheduling algorithm is adopted as follows to packet loss and packet delay exceeding the target bud- get. Although some packet loss occurs, the voice quality is (i) α (i) PMix (n) = Pr PPF (n)f (R y, Que s) , (1) not affected if the amount of packet loss is less than out- age threshold. To proceed with this work, we assume that where Pr is the level of priority which provides the high (i) the packet error rate (PER) is kept within 2%. Moreover, value to VoIP users and PPF (n) is the value of scheduling at least 97% of VoIP users in the downlink should meet the metric of user i calculated by PF scheduling scheme for BE above criterion . trafﬁc. The delay function f (·) can be designed by f (·) = (Que s)β /(R y)γ , where Que s is the size of VoIP packets 2.5. Scheduling strategy that must be scheduled at n-th TTI, R y is the remaining delay budget from the current n-th TTI to the delay bound The MAC-hs sublayer in the Node B handles scheduling at the MAC-hs scheduler. For BE service, f (·) = 1. α, β in the period of 2ms that is called transmission time interval and γ are appropriate weight factor for each one. 1 1.0 0.95 0.95 w/ UE Rx Diversity 0.9 0.9 Percentage of UEs with BLER<2% Percentage of UEs with BLER<2% 0.85 0.85 w/o UE Rx Diversity 0.8 0.8 0.75 0.75 0.7 0.7 0.65 0.65 0.6 Scheduler Delay Budget=90ms 0.6 Scheduler Delay Budget=100ms w/ FB0 Scheduler Delay Budget=110ms 0.55 w/ FB1 0.55 Scheduler Delay Budget=120ms w/ FB2 Scheduler Delay Budget=130ms 0.5 0.5 0 1 2 3 40 50 60 70 80 90 100 110 Number of Frame−Bundling Number of user Figure 4. Outage versus Number of FB for dif- Figure 5. Outage versus Number of user for ferent scheduler delay budget different value of FB missions, and the combined signal has higher probability 3. System-Level Simulation Conﬁguration of being successfully decoded. Finally, it has to be men- tioned that we simulate 100,000 TTI snapshots in average To investigate the capacity of VoIP service over HSDPA for investigating the performance of the system. The main system with a mixed voice and BE trafﬁc, a system-level simulation parameters are summarized in Table 2. computer simulation is accomplished in this paper. The simulations are carried out with a regular hexagonal 19 cel- 4. Simulation Results lular model, where the distance between Node B is 1km. Mobile terminals should be uniformly distributed on the 19- cell layout for each simulation run and assigned different In this section we evaluate the capacity of VoIP with the channel models according to the channel model assignment effect of FB and BE trafﬁc in the combination of various probability speciﬁed in . Note that a realistic model of fading channel environments (Ped.A 5% + Ped.B 45% + the wave propagation plays an important role for the signif- Veh.B 50%). As before described, capacity is deﬁned as icance of the simulation results. Shadowing is modelled by the number of UEs satisfying above outage condition of all a log-normal fading of the total received power and a ba- UEs, this is more than 97%. If an UE’s combined PER is sic attenuation is determined by the Hata model. Moreover, more than 2%, the user is considered in outage. The various we reserved the resources for control and common channel results are investigated by speciﬁc performance parameters overhead factors such as OVSF codes and HSDPA power such as the percentage of UEs satisfying outage limitation to obtain the provided simulation results. As mentioned and the sector throughput. above, we applied the RTP/UDP/IP packet header compres- sion using IETF RFC 3059 where the total size of all com- 4.1. The eﬀects of FB on the capacity of pressed header is with 3 bytes (1byte ROHC base header VoIP service over HSDPA + 2bytes UDP checksum). Further, the fast link adapta- tion and hybrid ARQ (HARQ) modules are employed in The effects of FB on the capacity of VoIP service without this work for throughput enhancement. While traditional considering for BE service are investigated in Fig. 2, Fig. 3, WCDMA systems use power control to mitigate channel Fig. 4 and Fig. 5. Fig. 2 shows the outage performance as a fading, HSDPA employs rate control based link adaptation function of scheduler delay budget for different number of where Node B transmits at full power and adjusts modu- user when there is no FB (FB0), with or without UE receiver lation and coding sets (MCS) according to channel varia- diversity. Here, UE receiver diversity means that the po- tions, maximizing instantaneous usage of the wireless chan- tential of achieving maximum-ratio combining (MRC) gain nel. HARQ scheme is also introduced to recover transmis- from the 2 receiver antenna in the UE-site. From the ﬁgure, sion failures. When mobile detects a transmission failure, it we observe that the employment of UE Rx diversity result sends a request to Node B for retransmission. Mobile com- in the additional capacity over the corresponding UE with bines soft signals of both original and subsequent retrans- single antenna system. This is because the 0 1 Percentage of VoIP UEs with BLER<2% 0.5 Total Throughput [Mbps] 0.8 1 0.6 0.4 1.5 0.2 2 110 100 0 90 80 110 Sc 80 70 100 hed 70 60 90 ule 60 50 80 80 rd 50 40 Sc 70 ela 40 30 hed 70 yb ule 60 ud 30 20 UEs r d 60 50 50 get 20 10 f VoIP ela 40 [m 10 be o yb 40 30 sec 0 Num ud get 30 20 20 UEs ] 10 f VoIP [m sec 10 be o 0 Num ] Figure 7. Total throughput versus Delay bud- Figure 6. Outage versus Delay budget and get and number of VoIP users number of VoIP users Table 4. Capacity of VoIP without FB in a Table 3. Capacity of VoIP with FB in a single mixed trafﬁc scenario trafﬁc scenario Capacity Ped.A Ped.B Veh.B Comb. Voice Capacity T-put[kbps] w/o UE Rx FB0 60 85 80 80 WCDMA AMR(CS) 66 406 Diversity FB1 40 80 80 75 (Rel.’99) VoIP(PS) 54 410 FB2 25 70 55 55 HSDPA Only VoIP 80 395 w/ UE Rx FB0 135 130 110 110 Mixed 65 589 Diversity FB1 125 125 105 105 Only BE - 1602 FB2 120 115 95 95 Remark w/o UE Rx diversity Remark 66 : CS Capacity on WCDMA 54 : VoIP on WCDMA than 110ms when FB0 . In summary, these results have shown that the FB can be employed in VoIP service by giv- probability of success for the initial transmission of VoIP ing larger scheduler delay budget since reduction of voice packet becomes more increase. In Fig. 3, we evaluate packet transmission delay can fully compensate the FB de- the effect of FB on outage performance, when the num- lay. Fig. 4 characterizes the effect of FB on outage perfor- ber of users is 60 in case of UE single antenna and 100 mance as a function of the number of FB when the number with receiver diversity. According to the results, the voice of user is 60 in case of UE single antenna. We note from transmission with FB has a higher sensitivity to the delay the ﬁgure that the outage performance is little degraded as budget in the scheduler. Especially, it is more intensive in a RF scheduler delay budget increases. Fig. 5 characterizes constraint of small delay latency, since FB increases VoIP the outage performance as a function of the number of users packet arrival interval and packet size. For example, when when FB=0, 1, and 2. Here, 90ms maximum delay budget aiming for an identical percentage value of 0.9, in case of is assumed. We note from the ﬁgure that for FB=1, the out- 100 UEs, required delay latency to achieve outage is 35ms, age performance is little degraded as the number of users 65ms, and 100ms for FB0, FB1, and FB2, respectively. This increases. By contrast, for FB=2, it did a great deal of per- imply that the VoIP capacity may be reduced if the FB is in- formance degradation to outage. Table 3 summarizes the cluded, due to the insufﬁciency of delay budget in RF sched- capacity of VoIP with FB or without FB in various prop- uler (110ms when FB0, 90ms when FB1, 70ms when FB2). agation conditions for multi-path fading environments for However, note that the VoIP capacity may not be reduced HSDPA. The results conﬁrm that VoIP service over HS- although the FB is employed, if the available delay budget DPA without FB provides signiﬁcantly higher capacity if in RF scheduler is larger compared to both VoIP and circuit-switched voice trafﬁc on WCDMA (Release’99) , . References  H. Holma and A. Toskala, WCDMA for UMTS, John 4.2. The capacity of VoIP service in a mixed Wiley and Sons, Third Edition, 2004 traﬃc scenarios  B. Douskalis, IP Telephone, Englewood Cliffs, NJ: Prentice-Hall, 2000 Fig. 6 and Fig. 7 present the capacity of VoIP service  W. Wang, S.C. Liew and Victor O.K. Li, ”Solutions to without considering for FB in a mixed voice and best-effort Performance Problems in VoIP Over a 802.11 Wireless trafﬁc scenario, where UE receiver diversity does not in- LAN,” IEEE Trans. Veh. Technol., vol.54, no.1, pp.366- cluded. Fig. 6 shows the outage performance as a function 384, January 2005 of scheduler delay budget and the number of VoIP users in the combination channel environments. We observe that  Q. Bi, P.C. Chen, Y. Yang and Q. Zhang, ”An Analysis VoIP service can be supported with satisfying the outage of VoIP Service Using 1xEV-DO Revision A System,” limitation if the delay budget of scheduler and the number IEEE JSAC, vol.24, no.1, pp.36-45, January 2006 of VoIP users are suitable. For example, when aiming for an outage percentage value of 0.97 that is the outage lim-  B. Wang, K.I. Pedersen, T.E. Kolding and P.E. Mo- itation to provide VoIP service, in the case of scheduler gensen, ”Performance of VoIP on HSDPA,” IEEE Proc. delay budget, the required delay to achieve outage is over VTC 2005, Spring, 2005 60ms. And in the case of the number of VoIP UE, the max-  3GPP TR 25.896, Feasibility Study for Enhanced Up- imum capacity user to provide voice service is below 65. link for UTRA FDD In Fig. 7, the effects of VoIP service on total throughput for the mixed trafﬁc scenarios are evaluated, when the num-  3GPP TS26.236, Packet switched conversational multi- ber of BE trafﬁc UEs is 40. According to the results, we media application; Transport protocols note that the increase of the number of VoIP users exhibits throughput worsen remarkably. This is because VoIP users  IETF RFC 3059, Attribute List Extension for the Ser- occupy an overwhelming majority in the given scheduling vice Location Protocol, February 2001 times to transmit packets by scheduler’s priority strategy.  G. Rittenhouse and H. Zheng, ”Providing VoIP Ser- Finally, table 4 summarizes the capacity of VoIP and the vice in UMTS-HSDPA with Frame Aggregation,” Proc. total throughput on HSDPA in combination channel condi- ICASSP, 2005 tions. VoIP service in mixed conditions has a lower user capacity than VoIP only service, although VoIP user is ser-  IETF RFC 4352, RTP Payload Format for the Ex- viced with higher priority by scheduling strategy. This is tended Adaptive Multi-Rate Wideband (AMR-WB+) due to the reserved resources for control and common chan- Audio Codec, January 2006 nel for BE trafﬁc users on HSDPA.  ITU-T G.114, Transmission systems and media - Gen- eral characteristics of international telephone connec- tions and international telephone circuits 5. Conclusion  3GPP TR 25.853, Delay Budget within the Access Stratum We have examined the system capacity of VoIP service  3GPP2 TSG-C.R1002-0, CDMA2000 Evaluation on HSDPA considering in the context of FB and BE trafﬁc Methodology, version 1.0, 2004 service. Our simulation results show that the capacity of VoIP trafﬁc is reduced because of FB and BE trafﬁc service,  3GPP TS25.101, User Equipment radio transmission and the throughput of BE trafﬁc is decreased as VoIP service and reception (FDD) is taken into account. from the viewpoint of FB, the FB can be employed in VoIP service by giving larger scheduler delay budget since reduction of voice packet transmission delay can fully compensate the FB delay. However, HSDPA is attractive for provision of VoIP service as compared to WCDMA (Release’99) if FB is limited in number according to the scheduler delay budget in a mixed voice and best- effort trafﬁc scenarios.
Pages to are hidden for
"VoIP Service on HSDPA in Mixed Traffic Scenarios"Please download to view full document