J. Cent. South Univ. Technol. (2009) 16: 0629-0634
DOI: 10.1007/s11771-009-0104-5
A random adaptive method to adjust MAC parameters in
IEEE802.11e WLAN
WANG Jian-xin(王建新), MAKFILE S, LI Jing(李 婧)
(School of Information Science and Engineering, Central South University, Changsha 410083, China)
Abstract: The IEEE 802.11e standard is proposed to provide QoS support in WLAN by providing prioritized differentiation of traffic. Since all the stations in the same priority access category (AC) have the same set of parameters, when the number of stations increases, the probability of different stations in the same AC choosing the same values will increase, which will result in collisions. Random adaptive MAC (medium access control) parameters scheme (RAMPS) is proposed, which uses random adaptive MAC differentiation parameters instead of the static ones used in the 802.11e standard. The performance of RAMPS is compared with that of enhanced distributed coordination access (EDCA) using NS2. The results show that RAMPS can reduce collision rate of the AC and improve the throughput by using adaptive random contention window size and inter-frame spacing values. RAMPS ensures that at any given time, several flows of the same priority have different MAC parameter values. By using the random offset for the inter-frame spacing value and the backoff time, RAMPS can provide intra-AC differentiation. The simulation results show that RAMPS outperforms EDCA in terms of both throughput and end-to-end delay irrespective of the traffic load.
Key words: wireless local area networks (WLAN); IEEE 802.11e; QoS; medium access control (MAC)
1 Introduction
In recent years, there has been a significant development in access networks, shifting from traditional wired network to wireless local networks (WLAN), which is more versatile, more inexpensive and easier to deploy. WLAN can now be found almost everywhere: in hotels, residential areas, university campuses, enterprises and in hotspot areas such as commercial areas and airports. The increase in popularity of WLAN has been accompanied by an equivalent increase in the demand for complex applications such as multimedia applications.
The Ethernet MAC protocol is broadly used in wired LAN. In the last few years, the IEEE 802.11 medium access protocol has received wide market acceptance and has become the de-facto standard for wireless LAN. The 802.11 MAC provides two access modes: distributed coordination function (DCF) and point coordination function (PCF). The PCF is designed to provide the quality of service provision for best-effort traffic. However, this QoS support is insufficient to accommodate stringent QoS requirements for time bound traffic such as the one used in multimedia applications [1-2]. For this reason, the IEEE 802.11 committee proposed a new standard called 802.11e [3].
The IEEE 802.11e enhances the legacy 802.11 access scheme by introducing a medium access method called the hybrid coordination function (HCF). HCF employs two different channel access schemes [4]: a contention-based channel access referred enhanced distributed coordination access (EDCA) and poll-based HCF controlled channel access (HCCA) [5]. For being widely deployed, the contention-based EDCA is focused on in this work. The main feature of EDCA is that it can provide prioritized QoS differentiation to different traffic classes based on values of contention window size (CW), arbitration inter-frame spacing (AIFS) and transmission opportunity (TXOP).
A significant number of studies have been carried out to investigate the efficiency and effectiveness of the EDCA function. Most of these studies indicated that under heavy traffic loads, the performance of EDCA declined significantly [6]. This is attributed to the fact that, EDCA uses static values for the priority parameters used for differentiating different traffic classes. Several methods were proposed to solve the static parameters problem mentioned above. Most of the proposed schemes aimed at adopting one or more priority parameters (CW, AIFS or TXOP) to the current state of the network. These schemes can be characterized as
either backoff-based priority schemes, IFS-based priority schemes or hybrid priority scheme [7]. For instance, in Refs.[6-7], adaptive enhanced distributed coordination function (EDCF) adaptively changes the contention window according to the node collision rate.
Although these schemes can enhance the quality of service of real-time traffic as well as the performance of EDCA, the performance levels gained are not optimal. The reason is that these schemes only provide differentiation between different traffic classes [8], but do not provide differentiation within the same traffic class. Thus, the collision rate remains significantly high during the high traffic loads.
In this work, by using the adaptive random MAC differentiation parameters, a scheme was proposed to reduce the intra-AC collisions and ultimately enhance the performance of EDCA.
2 IEEE 802.11e EDCA
As mentioned earlier, EDCA is the distributed contention-based access scheme of the IEEE 802.11e HCF mechanism that extends the legacy DCF access scheme. In EDCA, different traffic frames access to the shared channel in a differentiated manner based on their priorities as defined by the higher network layer. Each quality of service station (QSTA) in EDCA consists of four priority queues. Fig.1 shows access categories (ACs) into which frames from the higher layers are mapped according to their priorities. The ACs are allocated based on different EDCA contention parameters: CWmin(AC), CWmax(AC) and AIFS(AC) that are used to contend for a TXOP time [9]. TXOP is defined on per AC basis. It refers to the period in which a QSTA wins the chance to transmit data.
Fig.1 Schematic diagram of IEEE 802.11e access categories
The EDCA uses a carrier sense multiple access with collision avoidance (CSMA/CA) protocol similar to the one used in the legacy DCF [10]. This means that prior to a transmission, a QSTA that has frames to be transmitted, first senses the medium to determine whether the medium is busy or not. If the medium is idle, the QSTA will wait for a period equal to AIFS(AC) before transmission. On the other hand, if the medium is busy, the QSTA/AC will draw a pseudorandom number of slots called backoff timer that will be decreased or frozen when the medium becomes idle or busy, respectively. When the value of backoff timer reaches zero and the medium is idle for AIFS(AC), the QSTA will transmit the pending frames. The backoff time is selected randomly from [1, 1+CW(AC)], where CW(AC) is the contention window size belonging to a particular queue, and lies in the range [CWmin(AC), CWmax(AC)].
If the values of backoff timer belonging to different ACs of the same QSTA reach zero at the same time, a virtual collision will occur. In the event of the virtual collision, the AC with the highest priority is granted to have transmission opportunity and the other ACs perform a backoff with increased CW(i) values [11]. In addition to the virtual collision that occurs within a QSTA, packets belonging to different QSTAs may also collide during their transmission in the channel. In this case, the QSTAs whose packets have collided double their contention window sizes (CW(i)) and select a new backoff timer value. If packets do not collide during their transmission in the medium, the successful transmitters reset their contention window sizes to CWmin(AC).
CW(i)=min[CWmax(AC), 2CW(i)] (1)
In essence the backoff procedure operates in such a way that on average, an AC with the highest priority is assigned to the lowest backoff and AIFS values [12]. This ensures that the highest priority data frames have the shortest channel access delay. Therefore, they have a greater chance of accessing the channel than the lower priority frames [13]. Table 1 shows different standard EDCA parameters as defined for each AC. AC0 is the highest priority AC and AC3 is the lowest AC.
Table 1 Default MAC differentiation parameter sets
3 EDCA intra-AC differentiation problem
In EDCA, four ACs are allocated with different MAC QoS parameters in each QSTA, but ACs with the same priority have the same OoS parameters among different QSTAs, i.e. QSTA1 will have the same parameter of AC0, as those on QSTAi (i=2, …, n).
The use of the same static QoS parameters for the same priority AC on different stations introduces a problem in which the frames belonging to the same priority class are not differentiated from each other. The lack of differentiation causes collision among the multiple ACs of the same priority while they try to transmit packets simultaneously. When the number of stations in the network increases, the frequency of the collisions also increases, resulting in performance degradation. This problem is termed as the intra-AC differentiation problem and is explained further by using a simple hypothetical scenario.
Consider a situation in which an IEEE802.11e based WLAN has a single station. According to the standard, this station has four independent ACs, each with its own distinct set of MAC differentiation QoS parameters: CWmin(AC), CWmax(AC) and AIFS(AC). When the ACs in the station have data packets to send, they will compete for transmission opportunity using their unique EDCA MAC parameters. Because they use unique and independent parameters, their differentiation can be guaranteed in accordance with their priorities.
However, if the number of stations in the WLAN is increased to two or more, the competition for the shared medium will become fiercer because the ACs within individual stations will not only contend with each other, but contend with ACs from other stations. In each station, different ACs will contend for the medium using their distinct parameters as mentioned above. Whereas among different stations, all the corresponding ACs use the same EDCA MAC differentiation parameter set that can only take a limited number of values. Since all the stations have the same values for the same AC, these stations can be viewed as competitors for the values of a particular differentiation parameter. Therefore, increasing the number of the stations in the network can also be considered as increasing the number of stations that compete for specific values of a QoS parameter. It becomes obvious that the more saturated the network becomes, the more stations competing for limited values of the EDCA parameters there will be. This couples with the fact that EDCA does not provide the mechanism that one station can know about the condition of other stations. It seems that several of the same priority AC entities belonging to different stations choose the same backoff values. This inevitably results in collisions.
4 Adaptive random MAC parameters scheme
In this section, we discuss the proposed scheme referred to as random adaptive MAC parameter scheme (RAMPS), which aims at addressing the intra-AC differentiation problem discussed in the preceding section. In essence, RAMPS is an enhancement of the IEEE802.11e EDCA that aims at minimizing the high collision rate observed in EDCA at high traffic loads. RAMPS algorithm is based on random selection of two dynamic QoS parameters: CW and AIFS.
4.1 RAMPS algorithm
The CW and the AIFS play an important role in determining the time that a particular station has to wait for prior to a transmission. On average, an AC waits for a period which is equal to the backoff time plus the deferred time. The backoff time is based on selecting a random number from a range of values defined over the contention window size, whereas the deferred time is determined from the AIFS and is measured in timeslots. For the same AC on different quality stations, the contention window ranges and the AIFS values are the same. Therefore, it is possible for these same ACs to select the same values for the above parameters, which results in high collision rates especially when the channel load is very high.
We therefore introduce dynamic random offset values for both the CW and AIFS values in order to eradicate the above-mentioned problem.
4.2 Contention window based method
In this method, the random offset is determined based on the CW and calculated by taking into consideration the per-AC drop rate at each station.
The per-AC drop rate a(i) is calculated as a ratio of the number of packets dropped L to the number of packets sent S by an AC over a period of time. We use the drop rate because it shows the gap between the application bit rate and throughput of the physical medium, and it indicates the state of the network’s quality of service [14].
(2)
To adapt the value of CW according to the network conditions, using a(i), the maximum of the offset Fmax is calculated as
Fmax(i)=[CW(i)+1]a(i) (3)
In Eqn.(4), the actual offset value F(i) is selected randomly at an interval [0, Fmax(i)].
F(i)=random[0, Fmax(i)] (4)
The offset value is selected according to the network state to adjust the total backoff time. For instance, if the network is congested, F(i) will be large, which in turn increases the number of possible offset values. This reduces the likelihood of selecting the same offset value.
Then, the total backoff time T can be obtained as
T=random[1, CW(i)+1]+F(i) (5)
Adding the random offset value to the random backoff value increases the degree of randomness by a factor of [1+F(i), CW(i)+1+F(i)], and it also provides priority differentiation of traffic belonging to the same class as well as to different classes. Since there is little probability of which CW of higher class is higher than that of lower class, RAMPS can also provide good inter-AC differentiation.
4.3 Random AIFS-based method
In addition to the contention window size based intra-AC differentiation scheme described above, RAMPS uses an AIFS-based scheme to further achieve intra-AC discrimination of the same traffic on different QSTAs. This scheme is used in support of the CW-based scheme to ensure that two ACs of the same priority are unlikely to choose the same backoff value. Some differentiation will still be achieved through the use of AIFS. In the EDCA the AIFS is determined as follows:
AIFS(AC)=SIFS+AIFSN(AC)+SlotTime (6)
where the SIFS refers to the short inter frame spacing that is the shortest waiting time used for short messages, and AIFSN is arbitrary inter frame spacing number that is the number of time slots after the SIFS. A station has to defer the SIFS before either invoking a backoff or starting a transmission [15].
RAMPS extends the use of the AIFS to provide intra class differentiation in addition to per-AC differentiation of EDCA. This is achieved by allocating an interval of AIFSN values [N, M] to each AC instead of a fixed single value that is used in EDCA. The range of AIFSN in each AC is defined, and the value of AIFSN(i) is unique. AIFSN(i) is greater than AIFSN(i+1) ,where AIFSN(i) is the value for a higher priority AC and AIFSN(i+1) is that of the lower priority AC. Whenever an AC enters a deferred period, it will select a random value from the AIFSN range. This value is then added as an offset to the previous value of AIFSN as in Eqn.(7). The resulting value is then used as the current AIFNS value that is used for calculating the AIFS as in Eqn.(6).
AIFSNnew(AC)=AIFSNold(AC)+AIFSNoffset(AC) (7)
where AIFSNoffset(AC)=random[N, M].
In Eqn.(7), the random offset selected is always added to the previous AIFSN value (AIFSNold) to obtain the final offset. The current offset value that is added to the previous one increases the degree of randomness of the scheme. It is obvious that if the new AIFSN value is not limited, it will lead to infinitely high values. To avoid this we set the minimum and maximum limits to regulate the AIFSN value. The probability having two ACs with the same AIFSNold and AIFSNoffset is very small especially since both AIFSNold and AIFSNoffset are random numbers. Therefore, using both the old and the current offset values to determine the final AIFSN value guarantees that the final value is always unique.
5 Simulation and results
5.1 Overall performance
The performance of RAMPS is evaluated and compared with that of EDCA implementation using the network simulator (NS2) [16]. Fig.2 shows the topology used for the simulation.
Fig.2 Network topology for simulation
The topology is defined over a 670 m×670 m network range and is composed of several wireless nodes, each with four different priority flows from AC0 to AC3, where AC0 is the highest priority and AC3 is the lowest priority flow. The nodes in network send packets to access point using constants bit rate (CBR) traffic over user datagram protocol (UDP) at a data rate of 24 Mbit/s. Table 2 shows the offset range of AIFNS values.
Table 2 Offset range of AIFNS values
The overall throughput and delay are used as the performance measures, and the results are shown in Fig.3.
Fig.3(a) shows that RAMPS provides significantly more total throughput compared with EDCA. In fact, RAMPS achieves approximately 40% in throughput over EDCA. In Fig.3(a), the average throughput of EDCA is almost equal to that of RAMPS when the number of nodes is between 2 and 4. But as more nodes are added to the network the EDCA throughput declines significantly while that of RAMPS increases at a constant rate. While the throughput shows the network efficiency, the delay is another important performance indicator [17]. The delay of RAMPS compared to that of EDCA is shown in Fig.3(b). The delay of RAMPS is lower than that of EDCA. The reason is that in RAMPS the random adaptive offset values used ensure that at any given time each traffic has a distinct backoff value and thus reduces collisions. Reducing the number of collision implies less retransmission and packet drop. Most packets are delivered successfully so the delay becomes lower.
Fig.3 Performance comparison of RAMPS and EDCA: (a) Total throughput; (b) Delay
The throughput of the four different priority flows for RAMPS and EDCA is shown in Fig.4. In Fig.4(a), it is observed that the RAMPS scheme protects the throughput of all the flows while still maintaining the priority order of the flows. However, in EDCA (see Fig.4(b)) the throughput for the four traffic priorities declines with increasing number of stations. Moreover, the observed throughput for the two highest priority flows for RAMPS AC0 and AC1 is higher than that for EDCA.
5.2 Evaluation of intra-AC differentiation
To show the effectiveness of using random offset value in achieving finer priority differentiation of traffic we create a scenario in which only the highest traffic flows are present. In the scenario, three flows of the highest priority are created, and their performance is evaluated in the simulation environments, running from 20 s to 200 s. In order to simplify the simulation, the fixed values for the contention offset as well as the AIFSN values are shown in Table 3, and the observed results of the simulation are presented in Fig.5.
Fig.5 shows the average throughput plot for the three flows. The observed throughputs are proportional
Fig.4 Per-flow throughput in RAMPS (a) and EDCA (b)
Table 3 Contention window offset and AIFNS values
Fig.5 Intra-AC differentiation of high priority traffic
to the values of the contention window offset and AIFSN used. In particular, the flow with the lowest contention offset and AIFSN value shows the highest average throughput, while those with the highest values show lowest throughput. It is therefore obvious from Fig.5 that using different parameters for the same traffic class provides intra-AC differentiation. In this way, RAMPS decreases the collision between nodes and thus achieves higher throughput compared with EDCA.
6 Conclusions
(1) A new traffic differentiation scheme RAMPS is proposed for IEEE 802.11 MAC level QoS provisioning.
(2) RAMPS extends the EDCA scheme by using adaptive random CW and AIFS values to reduce the collision rate of the AC and improve the throughput. RAMPS ensures that at any given time, flows of the same priority can have different MAC parameter values to guarantee channel access at different time.
(3) The performance of RAMPS is compared with that of EDCA using NS2. The simulation results indicate that RAMPS outperforms EDCA not only throughput but end-to-end delay.
References
[1] WANG Jian-xin, DENG Shu-guang, CHEN Song-qiao, CHEN Jian-er. A quality of service routing protocol based on mobility prediction in mobile ad hoc networks [J]. Journal of Central South University of Technology, 2003, 10(1): 53-57.
[2] FRIKHA M, FATMA B S, MAALEJ L, TABBANA F. Enhancing IEEE 802.11e standard in congested environments [C]// Proceedings of the Advanced Inter Conference on Telecommunications and Inter Conference on Internet and Web Applications and Services (AICT-ICIW’06). Guadeloupe: IEEE Computer Society Press, 2006: 78-78.
[3] ANSEL P, NI Q, TURLETTI T. FHCF: A simple and efficient scheduling scheme for IEEE 802.11e wireless LAN [J]. Journal of Mobile Networks and Applications, 2006, 11(3): 391-403.
[4] KIM H, MARWITZ L, KIM D K. Dynamic offset contention window (DOCW) algorithm for wireless MAC in 802.11e based wireless home networks [J]. Journal of Mobile Communications, 2003, 252(1): 1-16.
[5] SIRIS V A, COURCOUBETIS C. Resource control for the EDCA and HCCA mechanisms in IEEE 802.11e networks [C]// Proceedings of the 4th International Symposium on Modeling and Optimization in Mobile, Ad Hoc and Wireless Networks. Trento: IEEE Computer Society Press, 2006: 1-6.
[6] ROMBHADI L, NI Q, TURLETTI T. Adaptive EDCF: Enhanced service differentiation for IEEE 802.11 wireless Ad Hoc networks [C]// Proceedings of the IEEE Wireless Communications and Networking (WCNC’ 03). Inria: IEEE Computer Society Press, 2003: 1373-1378.
[7] LIN W, WU J. Modified EDCF to improve the performance of IEEE 802.11e WLAN [J]. Journal of Computer Communication, 2007, 30(4): 841-848.
[8] XI W H, WHITLEY T, MUNRO A, BARTON M, KALESHI D, HEIDE G. Effectiveness of QoS provided by IEEE 802.11e for different traffic types [C]// Proceedings of the IEEE 62nd Vehicular Technology Conference (VTC’05). Dallas: IEEE Computer Society Press, 2005: 1132-1136.
[9] YAMANE M, TAGASHIRA S, FUJITA S. An efficient assignment of transmission opportunity in QoS guaranteed wireless LAN [C]// Proceedings of the 7th International Conference on Parallel and Distributed Computing, Applications and Technologies (PDCAT’ 06). Taipei, China: IEEE Computer Society Press, 2006: 105-108.
[10] BONONI L, BUDRIESI L, BLASI D, CACACE V, CASONE L, ROTDO S. A differentiated distributed coordination function MAC protocol for cluster-based wireless Ad-Hoc networks [C]// Proceedings of the ACM International Workshop on Performance Evaluation of Wireless Ad Hoc, Sensor, and Ubiquitous Networks. Venezia: ACM Press, 2004: 77-86.
[11] SUDHANSHU G, CLIFFORD T, TODOR C. Improved performance of CSMA/CA WLAN using random inter-frame spacing algorithm [C]// Proceedings of the International Conference on Wireless Communications and Mobile Computing (IWCMC’ 06). Vancouver: ACM Press, 2006: 407-412.
[12] RAZAFINDRALAMBO T, VALOIS F. Performance evaluation of backoff algorithms in 802.11 Ad Hoc networks [C]// International Workshop on Modeling Analysis and Simulation of Wireless and Mobile Systems. Torremolinos: ACM Press, 2006: 82-89.
[13] MALLI M, NI Q, TURLETTI T, BARAKAT C. Adaptive fair channel allocation for QoS enhancement in IEEE 802.11 wireless LANs [C]// Proceedings of the IEEE Conference on Communications(ICC’ 04). Inria: IEEE Computer Society Press, 2004: 3470-3475.
[14] KSENTINI A, NAIMI M, NAAFA A, GUEROUI M. Adaptive service differentiation for QoS provisioning in IEEE 802.11 wireless Ad Hoc network [C]// Proceedings of the ACM International Workshop on Performance Evaluation of Wireless Ad Hoc, Sensor, and Ubiquitous Networks. Venezia: ACM Press, 2004: 39-45.
[15] HAMIDIAN A, KORNER U. An enhancement to the IEEE 802.11e EDCA providing QoS guarantees [J]. Journal of Telecommon Systems, 2006, 31(2/3): 195-212.
[16] The VINT Project Group. The ns Manual[EB/OL]. [2008-04-21]. http://www.isi.edu/nsnam/ns/doc/ns_doc.pdf
[17] GAO Wen-yu, CHEN Song-qiao, WANG Jian-xin. End-to-end delay bound of packets [J]. Journal of Central South University: Science and Technology, 2006, 37(1): 135-140. (in Chinese)
(Edited by CHEN Wei-ping)
Foundation item: Project(60673164) supported by the National Natural Science Foundation of China; Project(06JJ10009) supported by the Natural Science Foundation of Hunan Province, China; Project(20060533057) supported by the Specialized Research Fund for the Doctoral Program of Higher Education of China; Project(2008CB317107) supported by the Major State Basic Research and Development Program of China; Project(NCET-05-0683) supported by the Program for New Century Excellent Talents in University
Received date: 2008-10-22; Accepted date: 2008-12-27
Corresponding author: WANG Jian-xin, Professor; Tel: +86-731-88830212; E-mail: jxwang@mail.csu.edu.cn
