Fair and Effective Transmissions in IEEE e WLAN

Transcription

1 GESTS Int l Trans. Computer Science and Engr., Vol.19, No.1 17 Fair and Effective Transmissions in IEEE e WLAN Seongcheol Kim 1 and Jinoo Joung 1 1 Software School, Sangmyung University, Seoul, Korea {sckim, Abstract. The goal of the Enhanced DCF (EDCF) is to provide a distributed access approach that can support service differentiation in IEEE e Wireless LAN. In this paper we propose a control mechanism to support fair and effective transmissions in IEEE e Wireless LAN. Since the proposed mechanism transmits packets using the link states and transmitted contents, higher transmission efficiency can be obtained by reducing unnecessary bandwidth usage. Fairness among mobile hosts also will be considered in this paper. The proposed algorithm uses short-term and long-term transmission times of each traffic classes to control their CW min for fairness. 1 Introduction With the tremendous growth of Internet applications over wireless networks, wireless access will be considered as an important communication path. As in wired networks wireless communication should be able to support quality-of-service (QoS), because many applications have QoS requirements like throughput, delay, and bandwidth. So the medium access control (MAC) protocol in wireless networks has to support an efficient scheme to provide the requirements of applications. The IEEE standard has been widely used in most wireless local-area networks (WLANs) products available in the market. The IEEE MAC specifies two different MAC schemes in WLANs: the contention-based Distributed Coordination Function (DCF) and the polling-based Point Coordination Function (PCF). The contention-based DCF mechanism is based on the Carrier Sense Medium Access with Collision Avoidance (CSMA/CA) protocol. In a network with CSMA/CA MAC protocol, a node first senses the channel to transmit packet whether the channel is in use. If the channel is sensed to be idle for an interval greater than the Distributed Inter-Frame Space (DIFS), the node is allowed to send packet. But, if the channel is sensed as busy, then the node defers packet transmission. The node then starts the backoff procedure by selecting a random backoff count. While the medium stays idle, the backoff counter is being decremented every slot time, and when the counter reaches zero, the node sends packets. For better transmission efficiency, many researches considering network traffic have been done[1,2,3,4,5,6,7,8,9]. Link adaptation mechanisms for IEEE WLAN were studied by many researchers. But these studies have some limitations. Some of

2 18 Fair and Effective Transmissions in IEEE e WLAN them use the transmission rate. But they use the rate in a manner that the system performance is not optimized. The reason is because that the rate is based on the result of timing function and the number of unsuccessful transmissions. So those mechanisms cannot react quickly when the wireless channel condition changes. Furthermore some studies assume some kind of link condition or network condition, or some studies require some change in the MAC layer. In [9] the authors propose a novel link adaptation algorithm, which aims to improve the system throughput by adapting the transmission rate to the current link condition. The suggested algorithm selects the best rate for a particular frame transmission based on the Received Signal Strength measurements. With the measurements senders select one of the transmission rates among 1, 2, 5.5, and 11 Mbps. With this mechanism higher throughput can be obtained. But this mechanism is not fit in the following case. For example, if bandwidth limited application traffic like MPEG 2 is to be sent with the mechanism. In that case even the required bandwidth cannot be provided, the sender will send its data with the predefined transmission rate. But the transmitted data will be discarded at the receiver, because insufficient data is not useful. This wastes link bandwidth consequently. The proposed mechanism in this paper considers this situation. In other word, if a node cannot get enough bandwidth through link adaptation, the node will defer packet transmission. The mechanism in [1] also considers the network conditions. It adjusts the Contention Window (CW) size according to traffic class and provides relative priorities. But the mechanism also has a problem, in which the performance of the mechanism is affected by parameters used. In this paper we propose an enhanced EDCF mechanism supporting fair and effective transmissions in IEEE e WLANs. The proposed mechanism is adaptive to the link conditions. If a traffic class can not get sufficient bandwidth due to the bad link conditions, the traffic class doesn't try to send its packet even though it has chance to use the medium. The fairness in this paper is based on the transmission time each node used, according to the long and short terms. The rest of this paper is organized as follows: Section 2 introduces the Enhanced DCF (EDCF) and the details of the proposed mechanism are presented in Section 3. Section 4 presents and discusses the simulation results, and the paper concludes with Section The IEEE e EDCF In IEEE e a new access method, Hybrid Coordination Function (HCF), is introduced, which combines functions from the DCF and PCF mechanism. The EDCF is the contention-based channel access protocol. The goal of this mechanism is to enhance the DCF access mechanism of IEEE 802,11 and to provide a distributed access that can support service differentiation. The mechanism provides up to eight types of Traffic Classes or categories (TCs). It assigns smaller CW (Contention Window) sizes for higher priority classes and larger sizes to low priority classes,

3 GESTS Int l Trans. Computer Science and Engr., Vol.19, No.1 19 giving high priority classes can transmit before the low priority ones. IEEE e also uses different inter frame space (IFS) according to the priority of each TC. In other words, Instead of DIFS used in DCF, an arbitration inter-frame space (AIFS) would be used for each TC. The AIFS for a given class should be a DIFS plus some time slots. The relationship of IFS is shown in Figure 1. The TC with the smallest AIFS has the highest priority. A big difference between DCF and EDCF is that when the medium is detected as idle for a period of AIFS, the backoff counter is reduced by one at the beginning of the last slot interval of the AIFS period. The EDCF operates only during the Contention Period (CP). Every station may implement up to eight transmission queues which can be thought of as virtual stations inside a station with different QoS parameters that determine their priorities. Each TC within the stations contends for a Transmission Opportunity (TXOP) independently. The backoff process is carried out individually after detecting if the medium is idle for a time equal to its AIFS. Fig. 1. The EDCF IFS relationship With EDCF, after each successful transmission of TC i, the corresponding CW i will be `set to CW min. If a transmission fails, then the CW i will be calculated as follows: CW = min{ CW, CW PF } i i,max i i Where PF i is the persistent factor of TC i. After waiting for AIFS i, each backoff timer is set to a random number from [1, CW i +1]. 3. The proposed Enhanced EDCF mechanism The idea of the proposed scheme in this paper is based on the following. The EDCF parameters such as (CW min, CW max, AIFS) are sent through beacon frame from AP (Access Point) to each station. And AP can change these parameters according to the network traffics dynamically. Each traffic class receiving these values from the AP will compete with each other as an independent EDCF sender. So if the link becomes idle, each traffic

4 20 Fair and Effective Transmissions in IEEE e WLAN class decreases their backoff counter and tries to send frame. The propose protocol operate as follows. 1) An AP transmits EDCF parameters and transmission rate according to link states to each station. All receivers calculate their transmission rate. Each station compares its estimated rate and the required QoS. 2) If the estimated bandwidth is less than the required bandwidth, the traffic class freezes their backoff counter even if the link is idle. The traffic class will join the competition again when enough bandwidth is provided. 3) Even if the backoff counter is 0 but the estimated bandwidth is not sufficient for required QoS, then the traffic class puts off its transmission. Since the proposed protocol considers network traffics and the applications requirements to send data, not only unnecessary bandwidth consumption can be reduced but also link utilization can be increased. 4. Fairness As discussed previously, DIFS, maximum packet size, and CW min, are used to provide service differentiation in the IEEE DCF and the IEEE e EDCF protocol. As in [10], if we use larger frame, then higher throughput can be obtained. But the results also say that using different size of frames, we cannot get delay differentiation. Furthermore, using different value of DIFS, we also cannot get enhanced throughput for TCP traffic. But throughput can be adjusted with different CW min, value. If a node increases its CW min, then the node will get reduced chance to transmit packet. So throughput will be decreased. Based on this fact, traffic classes that used more transmission time than allowed time should increase their CW min. On the other hand traffic classes that used less transmission time than allowed should decrease their CW min, to get chance to transmit packet. So fairness can be enhanced among the competing nodes. That is a basic idea of this paper to support fairness among traffic classes and to enhance throughput. To support fairness among the traffic classes, we use the same modeling in [11]. There are K different traffic classes and each traffic class requires different QoS. And we assume that K mobile hosts generate K traffic classes and a station k generates traffic class. The expected bandwidth for class-k stations is given as in [11]. ρ k = Lk p EI [ ] + EC [ ] + ES [ ] sk, (1) In the above equation, let E[I], E[C], E[S], be the sum of the expected length of all idle periods, all unsuccessful periods, and the expected length of a successful period,

5 GESTS Int l Trans. Computer Science and Engr., Vol.19, No.1 21 respectively. And let L k be the expected proportional of bits a class-k packet will be transmitted in a successful period. Finally, let P s,k be the probability of successful transmission of class k packet. So the total packet cycle will be the sum of E[I], E[C], and E[S]. We can get the maximum throughput when E[C]=0. This means that we can get the maximum throughput without collisions. As we know the average backoff time is given by [12]. CWmin EI [ ] = 2 (2) If we assume that IEEE a PHY, MAC parameters, and 16-QAM 3/4 PHY modulation are used, then our transmission rate will be 36 Mbps[6]. Figure 1 shows the channel time allocation in case of maximum throughput. Fig. 2. Ideal channel occupancy times In this case we get transmission throughput as follows. Tpkt _ trans T = 36Mbps ( 34μs+ 68μs+ 16μs+ 44μs+ T ) pkt _ trans (3) In the above equation is the packet transmission time. The equation says that the throughput increase as the packet size is increased. This is the same result as in [8]. In the propose protocol time window is used to measure the amount of time each traffic class used during the fixed successful transmission time. The used time of the each traffic class means that the wireless medium stay in busy state for data transmission. During this time transmission may successful or not. Since the EDCF mechanism uses the CSMA/CA medium access protocol, each station can detect whether the medium is in busy or not. If a wireless link is used for data transmission, NAV's of all stations will be set as busy. And each station can measure its successful transmission time through beacon interval. The adequate transmission time of each traffic class according to its priority can be obtained as follow. Let φ i be the weight allocated to the traffic class i and be one

6 22 Fair and Effective Transmissions in IEEE e WLAN transmission cycle. And we assume that φ 1 > φ 2 > φ 3 > > φ N is satisfied, then we get the following relationship for i and T C TC φ < i φ j (4) The higher priority traffic class gets more transmission time than the lower priority traffic class. If a traffic class used less transmission time than its allocated time, then the traffic class can use more time to transmit packet in the future. And a traffic class used more transmission time than its allocated time, then the traffic class should reduce its transmission time for fairness. But a time window can affect on determining whether a traffic class increases its transmission time or not. A traffic class may decide its transmission time be increased in short term time. But its transmission time may be fair in long-term time. Figure 2 shows the short-term and long-term average time a node used. When short-term transmission time is higher than that of long-term transmission time, the traffic class should reduce its transmission time. And when short-term transmission time is lower than that of longterm transmission time, the traffic class can increase its transmission time. Fig. 3. long-term and short-term average transmission time As we described before, throughput depends on CW min (i). So we can get fairness among traffic classes by adjusting their CW min (i) value according to their allocated transmission time. To support fairness among the traffic classes, the proposed algorithm operates as follow. - fair transmission time of each traffic class is T C φ i - each traffic class estimates its long term/short term time T C (i) - compare the long-term time T C,L to the short-term time T C,S 1) T C,L > T C,S In this case traffic class can increase its transmission time by decreasing CW min (i) value. 2) T C,L < T C,S

7 GESTS Int l Trans. Computer Science and Engr., Vol.19, No.1 23 In this case traffic class i should decrease its transmission time by increasing CW min (i) value. 5. Simulation Results To verify our protocol we used the same parameters in [1][8]. In the network model we assume that each node generates three traffics, High, Medium, and Low. These three traffics belong to the three classes of services: Audio (high priority), Video (medium priority), Background traffic(low priority). In the simulation the weight value for each class is chosen to be φ 1 : φ 2 : φ 3 = 1: 0.4: 0.2. And the ratio of the shotterm average and the long-term average is also chosen to be 1:100. The values are selected big enough to get reliable results. The network loads can be changed in this simulation as the number of nodes increase. All the results are averaged over 7 simulations. Figure 4 shows the medium utilization according to the number of nodes in AEDCF mechanism and in our mechanism. The proposed mechanism provides higher value in high network load. This is because each node sends its packet adequately in the proposed mechanism and doesn't try to send packet when it gets not enough bandwidth. Bandwidth will not be lost in the proposed mechanism as in the AEDCF mechanism. But two mechanisms show the same medium utilization in the low traffic loads. AEDCF Proposed Medium utilization[%] No. of Nodes Fig. 4. The comparison of link utilization As shown in Fig. 4, the delay of high class traffic of the proposed protocol is higher in the high load region than that of AEDCF mechanism. The reason is that as the number of nodes in the network increases, the allocated bandwidth to each traffic class will be decreased. So high class traffic may lose its chance to send its packet. If the high class traffic tries to send its packet in that case, the transmitted packet may useless at the receiver side. This makes bandwidth wastes. But in the low traffic region, the difference of delay between two mechanisms is trivial.

8 24 Fair and Effective Transmissions in IEEE e WLAN AEDCF Proposed Delay[ms] No. of Nodes Fig. 5. The delay of high class traffic Figure 6 shows the comparison of the low traffic throughput. When the number of nodes is more than 25, the proposed protocol provides better throughput of low traffic class. This is because when the network traffic is high, then the allocated bandwidth to the nodes is decreased. So high class traffic will get less chance to send its packets due to the insufficient bandwidth. On the other hand low traffic class gets more chance to transmit its packets. AEDCF Proposed Throughput[Kbps] No. of Nodes Fig. 6. The throughput of low class traffic 5. Conclusions In this paper we propose an enhanced EDCF mechanism supporting fair and effective transmissions in IEEE e WLANs. The proposed mechanism is adaptive to the link conditions. If a traffic class can not get sufficient bandwidth due to the bad link conditions, the traffic class doesn't try to send its packet even though it has chance to use the medium. In other words, the bandwidth-limited applications may not be transmitted in the proposed protocol when the link condition is bad. Avoiding the unnecessary transmissions can save bandwidth waste. But a high priority traffic class may put off its transmission according to the link states, the average delay of the traffic class may increase. And we consider supporting fairness among different kind of traffic classes. We use the long-term and short-term average transmission time

9 GESTS Int l Trans. Computer Science and Engr., Vol.19, No.1 25 each traffic classes used and the allocation times to traffic classes. According to the differences between these measured values and the allocation time, we adjust the CW min to provide fairness among the traffic classes. Our protocol shows good performances compared to the other studies. References [1] Lamia Romdhani, Qiang Ni, and Thierry Turletti, "Adaptive EDCF: Enhanced Service Differentiation for IEEE Wireless Ad-Hoc Networks," IEEE WCNC 2003 (Wireless Communications and Networking Conference), New Orleans, Louisiana, USA, March 16-20, [2] Daqing Gu, Jinyun Zhang, "QoS enhancement in IEEE wireless local area networks," IEEE Communications Magazine, Volume: 41 Issue: 6, Page(s): , June [3] Priyank Garg, Rushabh Doshi, Russell Greene, Mary Baker, Majid Malek and Maggie Cheng, "Using IEEE e MAC for QoS over Wireless," IPCCC'03, [4] Sunghyun Choi, Javier del Prado, Sai Shankar N, and Stefan Mangold, "IEEE e Contention-Based Channel Access (EDCF) Performance Evaluation, in Proc. IEEE ICC'03, Anchorage, Alaska, USA, May [5] Sunghyun Choi, "IEEE e MAC-Level FEC Performance Evaluation and Enhancement," in Proc. IEEE GLOBECOM'02, Taipei, Taiwan, November [6] D. Qiao and S. Choi, "Goodput Enhancement of IEEE a Wireless LAN via Link Adaptation", in Proc. IEEE International Conference on Communications ICC'01, Helsinki, Finland, June 11~14, [7] Grilo and M. Nunes, "Performance Evaluation of IEEE e," The 13th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, vol.1, pp , Sept [8] H. Zhu, G. Cao, A. Yener and A. D. Mathias "EDCF-DM: A Novel Enhanced Distributed Coordination Function for Wireless Ad Hoc Networks ", IEEE ICC, Paris, France, June [9] Javier del Prado Pavon and Sunghyun Choi, "Link Adaptation Strategy for IEEE WLAN via Received Signal Strength Measurement," IEEE ICC'03. Volume: 2, pp , May 2003 [10] Vasilos A. Siris and Matina Kavouridou, "Achieving Service Differentiation and High Utilization in IEEE ," Proc. of Personal Wireless Communications (PWC) 2003, Venice, Italy, September [11] Yu-Liang Kuo, Chi-Hung Lu at al, "An Admission Control Strategy for Differentiated Service in IEEE ," GLOBECOM IEEE Global Telecommunications Conference, no. 1, pp , Dec

10 26 Fair and Effective Transmissions in IEEE e WLAN [12] D.Qiao and K.G. Shin, "Achieving Efficient Channel Utilization and Weighted Fairness for Data Communications in IEEE WLAN under the DCF", in Proc.The Tenth International Workshop on Quality of Service (IWQoS'2002), May 15~17, Biography Seong Cheol Kim Address: 7 HonhJi Dong, Chongro-Ku, Seoul Korea, [ ] Ph.D in EE, Polytechnic University [ ] Samsung Electronics [1997-present] Sangmyung University Tel: sckim@smu.ac.kr Jinoo Joung Address: 7 HonhJi Dong, Chongro-Ku, Seoul Korea, [ ] Ph.D in EE, Polytechnic University [ ] Samsung Electronics [2005-present] Sangmyung University Tel: jjoung@smu.ac.kr