Simulation environment

We use NS-2 tool (version 2.29) [23] with ns-2 802.11 support [24]. The simulation environment consists of one QAP and a varying number of stations. The number of these QSTAs depends on the requirement of the simulation. All stations operate on IEEE 802.11a. The MAC and physical parameters are shown in Table 4.1.

Table 4.1 MAC and physical parameters Data rate 54 Mbps Basic rate 6 Mbps PHY header 20 us SIFS 16 us PIFS 25 us DIFS 34 us

We use three codec for VoIP, G.711 (160bytes payload, 20ms packet duration), G.723.1 (24bytes payload, 30ms packet duration) and GSM 06.10 (33bytes payload, 50ms packet duration), and three type of video transmission with different frequencies (10fps, 15fps and 30fps ) but the same frame size (1000bytes payload). In this simulation, the VoIP transmission consists of alternating talk-spurts and silence intervals. According to [19], we set the talk-spurts period with mean length of 7.24 sec and silence period with mean length of 5.69 second. All video transmissions are constant bit rate (CBR). FTP transmission is used to transfer packet with 1000 bytes payload.

We assumed that there are no hidden stations, thus RTS/CTS feature is turned on, and we compared our scheme with the round robin scheduler. The short interval is defined as 10ms.

Those stations start their voice or video transmissions randomly between 2 second and 3 second,

and all stations start FTP at 2 second. The simulation ended at 18sec. We analyze the simulation results just between 3 second and 18 second because the simulation network and transmission will be more stable during this period. Simulation results emphysize the comparison of the jitter deviation, the access delay of CBR and VBR traffic streams, the average throughput of total channel between our adaptive time-stamp scheme and the original round-robin polling scheme.

4.2 Numerical Results Analysis

In this chapter, we will compare our adaptive time-stamp polling (ATSP) scheme with the original round-robin polling scheme. Then we will discuss those simulation results and explain that our proposed scheme is better than the original polling scheme.

Table 4.2 Simulation result and comparison

Adaptive Time-Stamp Polling Scheme

3/3 5.36/0.52 1.2870 53.34 5.30/0.34 0.8271 3.35 0.44/0.41 1115.82 1172.51 6/6 5.42/0.73 1.7399 106.38 5.34/0.33 0.8005 5.71 0.47/0.58 1020.83 1132.92 9/9 5.60/1.10 2.6873 159.61 5.38/0.46 1.1445 9.90 0.58/1.08 922.59 1092.10 12/12 5.75/1.11 2.6838 211.73 5.48/0.66 1.5643 15.68 0.73/1.39 805.01 1032.42 15/15 6.08/1.65 4.0470 264.04 5.63/0.94 2.2652 21.02 0.85/1.66 700.39 985.45 18/18 6.16/1.70 4.0900 317.20 5.74/1.02 2.0226 25.74 1.04/2.25 605.47 948.41

Round-Robin Polling Scheme

3/3 14.50/5.74 15.1721 53.22 13.17/4.94 13.2185 3.35 0.48/0.63 1028.58 1085.15 6/6 15.20/5.96 15.3292 105.90 14.78/6.30 17.7249 5.70 0.67/1.77 855.86 967.46 9/9 15.10/5.90 15.5656 158.81 14.41/5.91 15.9606 9.88 0.99/2.91 682.10 850.79 12/12 15.40/5.99 15.6554 210.82 15.79/5.97 16.2959 15.66 1.55/4.32 495.44 721.92 15/15 15.37/6.18 16.3778 262.82 14.87/6.19 15.8832 21.01 2.83/7.92 325.91 609.74 18/18 15.68/6.37 16.5981 315.86 15.32/6.28 16.1110 25.73 6.07/16.52 171.74 513.33

We show our total simulation result first in Table 4.2 in order to present the perspective of those two quite different schemes. The first column is the number of CBR and VBR traffic streams, 3/3 means there are three CBR and three VBR traffic streams and so on. Then we present the average access delay and its standard deviation, the jitter deviation and throughput of those CBR, VBR and FTP traffic. The last column is the throughout of the channel.

In the following sections, we discuss those simulation results of three aforementioned criteria separately, these include average throughputs, jitter deviations and access delays.

4.2.1 Throughput Enhancement

54Mb 802.11a with different polling scheme

500 600 700 800 900 1000 1100 1200

3/3 6/6 9/9 12/12 15/15 18/18

Number of CBR/VBR

Throughput (KB/s)

ATSP RR

Fig. 4.1 Average throughput against QSTA number for ATSP and RR schemes

Figure 4.1 shows the average throughout between 3sec and 18sec in different network conditions. When more and more QoS traffic streams appears on the same channel, it will cost longer time to access this channel, it will also waste more time on polling overhead, and FTP

traffic streams would have less time to contend for the channel. So the average throughput will decrease with the growth of the QoS traffic streams.

We can also notice a more gradual decrease in the average throughput for the adaptive time-stamp polling scheme comparing to round-robin scheme from Figure 4.1. RR scheme starts at 1085 .15 KB/sec and ATSP scheme starts at 1172.51 KB/sec when the numbers of CBR and VBR traffic streams are three. However, the gap between ATSP and RR schemes is getting larger when the number of QoS stations increases. The reason is that the throughputs of CBR and VBR traffic streams are almost the same for both schemes but the throughputs of FTP traffic streams are completely different when the network size is growing up (see Table 4.3). Therefore, ATSP scheme ends at 948.41 KB/sec and RR scheme ends at only 513.33 KB/sec when the numbers of CBR and VBR traffic streams are eighteen. From Figure 4.1 and Table 4.3, while both numbers of CBR and VBR traffic streams increase, ATSP scheme achieves a much better performance than RR scheme on the average throughput.

Table 4.3 Average throughput for different polling schemes Average Throughput (KB/sec)

CBR/VBR

ASTP_FTP RR_FTP ASTP_CBR RR_CBR ASTP_VBR RR_VBR

3/3 1115.82 1028.58 53.34 53.22 3.35 3.35

6/6 1020.83 855.86 106.38 105.90 5.71 5.70

9/9 922.59 682.10 159.61 158.81 9.90 9.88

12/12 805.01 495.44 211.73 210.82 15.68 15.66

15/15 700.39 325.91 264.04 262.82 21.02 21.01

18/18 605.47 171.74 317.20 315.86 25.74 25.73

4.2.2 Jitter Deviation Reduction

3/3 6/6 9/9 12/12 15/15 18/18

Number of CBR/VBR

Fig. 4.2 Standard deviation of jitter against number of QSTA

From Figure 4.2, we see the standard deviation of jitter in different numbers of CBR and VBR traffic streams. The jitters are calculated by subtracting previous transmission time from current transmission time of the same QoS traffic stream between the AP and the stations.

Generally, we can notice that the jitter standard deviations increase when there are more and more QoS traffic streams on this network just because heavy traffic load will deteriorate the stability of the network.

Figure 4.3 shows the jitter of one CBR traffic stream in RR polling scheme. Figure 4.4 shows the jitter of the same traffic stream in our ATSP scheme. Both are in the networks with eighteen CBR and eighteen VBR traffic streams. We can notice that the variations of jitter in RR polling scheme are relatively large, and the maximum jitter is higher than 30ms (see Figure 4.3).

Hence, the standard deviation of jitter for RR scheme will be large certainly. On the contrary, the line in the Figure 4.4 vibrates slightly, and the values are almost restricted within 5ms. This means the jitters of our ATSP scheme are almost negligible, and the simulation result and

analysis by Equation 3.9 are matched neatly.

From Figure 4.2, 4.3 and 4.4, ATSP scheme shows a much better jitter reduction than RR scheme regardless of the network size and the numbers of CBR and VBR traffic streams.

Jitter of one CBR traffic for RR scheme in 18/18 network

-40

Jitter of one CBR traffic for ATSP scheme in 18/18 network

-40

Fig. 4.4 Jitter between packets in ATSP scheme

4.2.3 Access Delay Improvement

Figure 4.5 shows the average access delay against different network sizes for different types of traffic stream and different polling schemes. This figure shows significant gaps between RR scheme and ATSP scheme regardless of the types of traffic streams. The higher delay in RR

scheme is due to the polling overheads and inaccurate polling time. While ATSP scheme shows a lower delay comparing to the RR schemes, this is because we can decrease our access delay by reducing the polling overhead and using the short interval polling function. From Figure 4.5, ATSP scheme presents lower access delay than RR scheme regardless of the network size.

Average Access Delay

0 2 4 6 8 10 12 14 16 18

3/3 6/6 9/9 12/12 15/15 18/18

Number of CBR/VBR

Average Access Delay (ms)

RR_CBR RR_VBR ATSP_CBR ATSP_VBR

Fig. 4.5 Average access delay against number of QSTA

Chapter 5 Conclusion and Future Works

In order to reduce the polling overhead and provide higher QoS for real-time traffic streams, we proposed a new polling scheme, called Adaptive Time-Stamp Polling (ATSP) scheme. In our proposed scheme, it contains three enhancement parts; one is the adaptive time-stamp poll scheduling which is responsible for calculating the next polling time of each traffic stream.

Basically, it calculates by adding the maximum service interval which is registered in the corresponding TSPEC to the current polling time, instead of adding the same service interval for all traffic streams. This method is more flexible and able to reduce the number of unnecessary polling and lessen the jitter deviation.

The second part is the short interval polling function which focuses on the access delay reduction by using the personalized short interval to poll some traffic stream between receiving first QoS data frame and second QoS data frame. During short interval polling periods, it calculates next polling time by just adding the short interval to detect a more accurate polling time. The last part is used to solve talk-spurt and silence alternation problem of voice traffic streams such as VoIP, it detects the silence traffic streams by counting the consecutive QoS-Null replies up to three, and considers them as in the talk-spurt state when receiving QoS data. When it considers that some traffic stream is in silence state, it polls this stream with a longer interval to reduce the overhead.

We use NS-2 tool (version 2.29) with ns-2 802.11 support which contains 802.11e module (both HCCA and EDCA) to simulate our adaptive time-stamp polling scheme and compare it with round-robin polling scheme. The results show that ATSP has significant improvement in terms of throughput, access delay and jitter deviation. The standard deviation of jitter for real-time traffic in ATSP scheme reduces by more than 60% comparing to RR scheme. The average delay in ATSP scheme also decreased more than 50% comparing to RR scheme and the

standard deviation of access delay is very little, this means frames will unlikely experience high access delay with our ATSP scheme. The most important thing is that we would not sacrifice the best-effort transmission for real-time applications. Best-effort traffic would not suffer from starvation when there are more and more QoS traffic streams added to the channel, and the total utilization of channel will be also improved.

The complexity of our ATSP scheme is a big issue because the HC needs to calculate polling time of each traffic stream respectively and observe which traffic stream has been in the silence state or during the short interval polling period. The short interval polling period of the traffic stream is much shorter than the whole real-time transmission and we use simple method to detect silence streams, so the HC won’t spend much time to observe those traffic streams. Truly complex part is to calculate polling time for each traffic stream. The more QoS traffic streams, the more the calculations. However, those calculations are almost simple additions, so we could conclude that ATSP scheme achieves a significant improvement.

In the future, we will improve our silence detection method to make those decisions more reliable, and find a more efficient mechanism about access delay reduction to minimize extra polling. Besides, the Direct Link Setup (DLS) and Block Acknowledge will be taken into account to support a more reliable and effective transmission for real-time applications on wireless LANs.

References

[1] IEEE Std 802.11e-2005.

[2] Ping Wavg, Hai Jiang, Weihua Zhuang, “IEEE 802.11E Enhancement for Voice Service”, IEEE Wireless Communications, February 2006.

[3] Q. Ni, “Performance Analysis and Enhancements for IEEE 802.11e Wireless Network”, IEEE Network, Vol. 19, pp.21-27, July-Aug. 2005.

[4] Ming-Chuan Hsu; Yaw-Chung Chen, “Enhanced PCF Protocols for Real-time Multimedia Services over 802.11 Wireless Networks”, Distributed Computing Systems Workshops, 2006. ICDCS Workshops 2006. 26th IEEE International Conference on, pp.56 – 56, 04-07 July 2006.

[5] Minoru Matsumoto and Tadashi Itoh, “QoS-guarantee Method for Public Wireless LAN Access Environments”, 2005 International Conference on Wireless Networks, Communications and Mobile Computing.

[6] Jungbo Son, Hosuk Choi ,Sin-Chong Park, “An Effective Polling MAC Scheme for IEEE 802.11e”, international Symposium on Communications and Information Technologies 2004 ( lSCIT 2004), Sapporo. Japan, October 26 - 29, 2004.

[7] Jun Zheng, Emma Regentova, “An Improved Polling Scheme for Voice Support in IEEE 802.11Wireless Network”, Proceedings of the International Conference on Information Technology (ITCC’05).

[8] Ray Y. W. Lam and Victor C. M. Leung, “Polling-Based Protocols for Packet Voice Transport Over IEEE 802.11 Wireless Local Area Networks”, IEEE Wireless Communications, February 2006.

[9] Xiyan Ma, Yanfeng Zhu and Zhisheng Niu, “Dynamic polling management for QoS differentiation in IEEE 802.11e wireless LANs”, 10th Asia-Pacific Conference on Communications and the 5th International Symposium on Multi-Dimensional Mobile, vol.1, pp.152-156, 2004.

[10] Rongbo Zhu and Yuhang Yang, “Adaptive Scheduler to Improve QoS in IEEE 802.11e Wireless LANs”, First International Conference on Innovative Computing, Information and Control (ICICIC '06), vol. 1, pp.377-380, 2006.

[11] Ping-Chi Wang, Kuochen Wang, Lung-Sheng Lee, “A QoS scheme for digital home applications in IEEE 802.11e wireless LANs”, IEEE 16th International Symposium on Personal, Indoor and Mobile Radio Communications, Vol. 3, pp. 1845-1849, 2005.

[12] T. D. Lagkas and G. I. Papadimitriou, “Priority Oriented Adaptive Polling for wireless LANs”, 11^th IEEE Symposium on Computers and Communication (ISCC’06), 2006.

[13] Xian Ma, Cheng Du, Zhishing Niu, “Adaptive Polling List Arrangement Scheme for Voice Transmission with PCF in Wireless LANs”, 10’th Asia-Pacific Conference on Communications and 5th International Symposiumon Multi-Directional Mobile Communication.

[14] Bandinelli. M, Chifi. F, Fantacci. R, Tarchi. D, Vannuccini. G, “A link adaptation strategy for QoS support in IEEE 802.11e-based WLANs”, IEEE Wireless Communications and Networking Conference, Vol.1, pp. 120-125, 2005.

[15] Hyun-Jin Lee and Jae-Hyun Kim, “A optimal CF-poll piggyback scheme in IEEE 802.11e HCCA”, The 8th International Conference Advanced Communication Technology, Vol.

3, pp.6 ,Feb. 2006.

[16] Il-Gu Lee, Jung-Bo Son, Sung-Rok Yoon and Sin-Chong Park, “Efficient block size based polling scheme for IEEE 802.11e wireless LANs”, 2005 IEEE 61st Vehicular Technology Conference, vol. 5, pp. 2869-2873, 2005.

[17] Fang-Yie Leu, Ching-Chien Kuan, Dr-Jiunn Deng, “A QoS provision multipolling mechanism for IEEE 802.11e standard”, International Conference on Wireless Networks, Communications and Mobile Computing, vol.1, pp. 392-397, 2550

[18] Paul T. Brady, “A model for generating ON-OFF speech patterns in two-way conversations”, Bell System Technical Journal, Vol. 48, pp. 2445-2472, Sept 1969.

[19] Shuang Deng, “Traffic characteristics of packet voice”, IEEE International Conference on Communications, Vol. 3, pp. 1369-1374, 1995.

[20] J.Naoum-Sawaya, B. Ghaddar, S. Khawam, H. Safa, H. Artail and Z. Dawy, “Adaptive Approach for QoS Support in IEEE 802.11e Wireless LAN”, IEEE WMob’05, Vol. 2, pp/

167-173, Aug. 2005.

[21] T. Raimondi and M. Davis, “Design Rules for a Class-based Differentiated Service QoS Scheme in IEEE 802.11e Wireless LANs”, ACM MSWiM’04, Oct. 2004.

[22] L. Romdhani, Q. Ni and T. Turletti, “Adaptive EDCF: Enhanced Service Differentiation for

IEEE 802.11 Wireless Ad-Hoc Networks”, IEEE WCNC, Vol. 2, pp. 1373-1378, Mar. 2003.

[23] http://www.isi.edu/nsnam/ns/

[24] http://yans.inria.fr/ns-2-80211/