OFDMA-Based Medium Access Control for Next-Generation WLANs
© H. M. Alnuweiri et al. 2009
Received: 4 August 2008
Accepted: 18 January 2009
Published: 23 February 2009
Existing medium access control (MAC) schemes for wireless local area networks (WLANs) have been shown to lack scalability in crowded networks and can suffer from widely varying delays rendering them unsuited to delay sensitive applications, such as voice and video communications. These deficiencies are mainly due to the use of random multiple access techniques in the MAC layer. The design of these techniques is highly linked to the choice of the underlying physical (PHY) layer technology. The advent of new PHY schemes that are based on orthogonal frequency division multiple access (OFDMA) provides new opportunities for devising more efficient MAC protocols. We propose a new adaptive MAC design based on OFDMA technology. The design uses OFDMA to reduce collision during transmission request phases and makes channel access more predictable. To improve throughput, we combine the OFDMA access with a carrier sense multiple access (CSMA) scheme. Data transmission opportunities are assigned through an access point that can schedule traffic streams in both time and frequency (subchannels) domains. We demonstrate the effectiveness of the proposed MAC and compare it to existing mechanisms through simulation and by deriving an analytical model for the operation of the MAC in saturation mode.
The throughput and scalability of a wireless local area network (WLAN) are greatly dependent on its medium access control (MAC) protocol and its associated multiple access (MA) scheme. The most important aspect of a MAC protocol is the ability to efficiently coordinate access among multiple contending stations (STA) for a shared resource. MAC protocols for WLANs must also address several other key issues, such as coping with a varying and potentially large number of users, maintaining acceptable delays for multimedia traffic, transmitting over a randomly varying wireless channel, and introducing minimal overhead to the transmitted data stream.
The original designs of MAC protocols for WLANs included the CSMA/CA (CSMA with collision avoidance) mechanism in IEEE 802.11 WLANs and time division multiple access (TDMA) in HIPERLAN [1–3]. The latter was never really accepted by the industry, while IEEE 802.11 (dubbed WiFi) quickly became the technology of choice for WLANs worldwide. TDMA required complex scheduling in a central node and used fixed time slot-based transmissions, as opposed to variable length packet transmissions. Such a scheme had limited flexibility in low cost WLANs that are dominated by bursty data traffic. Instead, simpler random access CSMA-based protocols were preferred. However, as the experience with deploying 802.11 based networks matured, it became obvious that the simple CSMA/CA MAC mechanism can become quite inefficient, mostly due to its collision prone random access and recovery processes. Although mechanisms to control and minimize the impact of collisions (such as collision detection in the CSMA/CD mechanism of Ethernet) exist, a wireless medium does not allow for collision detection through power measurement. Furthermore, the collision avoidance mechanism (CA), used in CSMA/CA protocols, has a very limited capacity in improving the performance of the MAC. In fact, the performance of the CSMA/CA protocol used in IEEE 802.11 standard deteriorates rapidly as the number of stations increases and/or the amount of traffic grows .
One of the main objectives of our proposed MAC is to address this issue of low throughput (or efficiency) of the MAC layer under heavy loading conditions in crowded WLANs. Another major issue is to improve support for persistent real-time traffic, produced by multimedia applications, while maintaining the random access and variable length packet transmission features of the original MAC. To achieve these goals, we utilize the orthogonal frequency division multiple access (OFDMA) scheme [5, 6]. OFDMA is currently used in the IEEE 802.16e standard for wireless metropolitan area networks . Multimedia traffic has traditionally suited more centrally controlled MAC protocols such as TDMA, over a random access network . In our proposed MAC, we utilize controlled CSMA access to provide better scheduling opportunity for multimedia traffic. At the same time, we preserve and increase the efficiency of the random access feature through combining OFDMA with CSMA.
In the next section, we briefly review some relevant works and techniques and elaborate on the advantages of an OFDMA-based MAC. We present our OFDMA-based MAC design in Section 3. The proposed design is then analyzed and evaluated in Section 4. Open issues and future research directions are discussed in Section 5.
2. Multiple Access Scheme for WLANs
Providing contention minimized random access is one of the main aspects of our proposed design. For this purpose, we can combine CSMA with any multiple access scheme such as frequency division multiple access (FDMA), code division multiple access (CDMA) (similar to the method in ), TDMA, or OFDMA.
It is well known that OFDMA has a clear capacity advantage over FDMA, therefore we do not discuss FDMA here. Also FDMA-based schemes such as multichannel slotted Aloha have been well examined in the past . As described below OFDMA has several advantages over CDMA and TDMA schemes as well and is our primary choice.
In CDMA technique, the transmitted signal is multiplexed with a high bit rate pseudocode sequence, to distinguish it from other simultaneous transmissions in time and frequency. However, this high bit rate transmission is susceptible to channel frequency selectively and requires complex channel equalization at the receiver to recover from the resulting intersymbol interference. It is widely known that indoor radio channels experience high frequency selectively due to their rich scattering environments. Furthermore, CDMA schemes suffer from the near-far phenomenon; a nearby STA's high power transmission may completely block out faraway STA's lower power transmissions. OFDM-based techniques such as OFDMA have been proven to perform well and are more easily implemented for WLANs.
TDMA-based MAC designs are known to be mostly suitable for constant bit rate voice communications but not for bursty data or video traffic . Moreover, in TDMA, where STAs share the available bandwidth in time, uplink connection request transmissions have to happen in exact time slots. Such a mechanism has strict synchronization that needs and requires all STAs to be aware of these time slots and to be perfectly aligned with the central controller. If CSMA rules are used with TDMA to avoid collision and relax synchronization requirements, the interframe spacing required between slots will reduce the throughput of the system.
OFDMA, on the other hand, is based on OFDM, which is inherently more suited for wireless channels. OFDM can convert the frequency selective channel into a parallel but orthogonal set of smaller frequency-flat channels which the receiver can deal with without needing expensive channel equalization. With the use of guard intervals, ISI can be completely avoided in OFDM systems. OFDMA allows assigning these subchannels to different STAs. Given this ability of OFDMA and its superior performance in wireless environment typical to WLANs, a system using OFDMA is of more interest to future multiple access wireless networks.
Some recent works on MAC design based on OFDMA or OFDM are reported in [11, 12]. The design in is based on a combination of OFDM and TDMA and attempts to provide better performance through intelligent assignment of subcarriers in time domain. This scheme is not designed for efficient random access which is one of our objectives. The work in presents a MAC that relies on OFDMA for all of its operation and resolves collisions through changing subchannel assignments. While such methods and the standard 802.16 MAC protocol provide novel schemes using OFDMA, they require complex scheduling, are more suitable for consistent voice-like traffic, and do not provide efficient random access capabilities. We present a new design that is different in many aspects and is more suitable for heterogeneous traffic, which is typical for WLANs. Our design is also more flexible and easily extendable for operation in different environments.
2.1. OFDMA System Specifications
An important aspect of an OFDMA system that is of interest to the design of a MAC scheme is the number of subchannels. OFDMA allows subcarriers to be grouped and assigned to different users. These groups of subcarriers are known as subchannels (also called subbands in some literature). In our scheme, the number of subchannels is primarily determined by the MAC to achieve optimal performance. This issue is discussed in Section 4.
The number of subcarriers in an OFDMA scheme is a design factor that can be adjusted for different environments. For IEEE 802.16 standard, a fixed separation of subcarriers is considered [7, 13]. Depending on channel bandwidth, which may range from 1.25 to 40 MHz, the number of subcarriers ranges from 128 to 2048. The 802.11 a/g standard uses OFDM with 64 subcarriers in a 20 MHz BW . We do not make an assumption on the number of subcarriers in this article.
Our scheme does not require any specific choice of guard interval for OFDM symbols, or the guard bands for the channel. Any existing PHY specification for OFDMA or OFDM systems (e.g, from 802.16 or 802.11 standards) is acceptable. Subchannels formed from grouping subcarriers together may be adjacent or may be distributed. In general, distributed permutations outperform adjacent subchannels in high mobility applications while adjacent subcarriers can be used to provide higher throughput in fixed, portable, and low mobility environments .
3. Hybrid OFDMA/CSMA MAC Design
Our proposed MAC uses a two-stage frame delivery process, consisting of a transmission opportunity request (TR) phase, followed by a scheduled data transmission (ST) phase. A station willing to send data has to inform the central controller, the access point (AP), of its required transmission opportunity (TO). The TO request is sent in a contention phase with reduced collision (using OFDMA). Once the AP receives this message, it will assign contention free TOs to the station. A station can also transmit a more elaborate TO request, such as, a periodic TO assignment for its voice or video traffic. The AP can then regularly assign TOs to the station, without needing subsequent TO requests from the station (similar to unsolicited access grant in 802.16 ). Apart from the obvious decrease in control messages that need to be sent, such a mechanism reduces contention-based TO requests, thereby providing additional QoS guarantees for multimedia traffic. Following the grant of a contention free TO, the station can send longer data frames.
The TR phase starts after an MIFS interval, following an explicit TR start message from the AP or automatically after sensing the entire channel (not just the subchannel) idle for RIFS seconds. Transition between TR and ST phases happens after sensing the entire channel idle for CIFS after the backoff slots in the TR phase. If there has been no message in the TR phase, and AP has no data or message to send, the TR phase is repeated with stations starting to count backoff slots after a RIFS again. In our scheme, the TR phase always uses OFDMA, whereby a set of subcarriers (i.e., a subchannel) is assigned to each station for sending the TR messages. When the number of subchannels is equal to or more than the number of stations, the TR phase will be contention free. To increase the utilization and the multiplexing gain for the TR phase, we can assign more than one station to each subchannel; this means that a probability of collision between stations exists within each subchannel. However, this probability can be capped through dynamically adjusting the number of subchannels and the number of stations assigned to each subchannel. A CSMA/CA scheme is used to resolve the probable collisions, and a limited number of time slots are allocated for the backoff process. This constitutes a hybrid and dynamic OFDMA/CSMA scheme for the TR phase. We evaluate the operation of different TR phase settings later in this article.
When the TR phase completes, the AP processes the received TRs and schedules the subsequent ST phase. Transmission in the ST phase is contention free and can be from one station to the AP or to another station. The contention free transmission may be done in several ways. A simple method is to assign all subcarriers to one station and schedule stations in the time domain (i.e., OFDM with time domain multiplexing). In this mode, either a transmission schedule (TS) for stations is broadcasted, or individual TO assignment messages are sent to stations at appropriate times. The other option for the ST phase is to utilize OFDMA to allow concurrent data transmission by multiple stations. In this mode, a subchannel assignment (SA) map is broadcasted to inform the stations of their assigned subchannels. The SA map may include downlink transmissions by the AP, and therefore it is not equivalent to the uplink map (UL Map) in other standards. Utilizing OFDMA in the ST phase incurs great complexity in scheduling stations and may lead to inefficiency if a schedule that fills all subchannels at all times is not feasible. In this article, we only consider the former case where OFDMA is used in the TR phase and the ST phase uses controlled CSMA.
3.1. To Request (TR) Phase
During the TR phase, each station can transmit its TO requests in its assigned subchannel according to the CSMA/CA rules. Each subchannel runs its own separate CSMA/CA procedure. An important part of the design for our system is the subchannel assignment for random access messages, that is, TO requests. Assuming that subcarriers (e.g., = 256) are available, and there are stations, we divide the channel to subchannels, or in other words, assign subcarriers to each station. If = 1 or , we have an all OFDM system with CSMA/CA operation, similar to 802.11. If , we have the other extreme ( ), which is an OFDMA system without contention (no need for CSMA). A balance can be found between the wasted bandwidth due to collision and that due to subchannels sometimes remaining unused. This is discussed in more detail in the next section.
To handle the possible contention in each subchannel, we employ a CSMA/CA mechanism. Each subchannel uses its own separate CSMA/CA procedure. Collisions are resolved using a random exponential backoff process. There are other collision resolution mechanisms such as changing subchannel assignment as described in . However, such methods increase system complexity without providing significant gain in performance. Thus we use independent CSMA/CA mechanisms in each subchannel. The CSMA/CA scheme operates similarly to the process used in 802.11, except for the carrier sensing in each subchannel that uses OFDM symbol detection instead of simple channel power sensing.
The TR phase timeline includes a limited number (denoted as ) of minimum length time slots followed by a transmission opportunity for the TR message. The minimum time slots are used by the backoff process of the CSMA/CA scheme to avoid collision. The duration of such time slots is derived from the minimum time that is needed for the detection of an OFDMA symbol in the subchannel. When a station wants to transmit a TR message, it chooses a random backoff number from the interval (0, ). The stations decrement their backoff counter with each empty slot in the TR phase passing; when the counter reaches zero, they transmit the TR message immediately if they sense no other station already transmitting in the subchannel. After a station transmits a TR message, it will wait for a response from the AP, in the form of a poll or position in the schedule. If no response is received, the station will interpret this as a collision (or lost packet), double its contention window size, and then select a new random backoff number from the new contention window. The lower and upper limits of the contention window size can be traffic class dependent to provide prioritized quality of service. These values can also be adaptively assigned to stations by the AP, to achieve better performance.
To achieve even higher performance, the AP can dynamically change the number of subchannels and assign stations to different subchannels. This can be done using the network allocation vector (NAV) field in the standard CTS frame sent by the AP. The access point keeps track of the number of associated stations ( ) and the number of active stations ( ). Active stations are ones that have set up a traffic stream with the AP, have indicated a nonzero queue size, or have transmitted in the past beacon intervals ( being a configuration parameter). When assigning stations to subchannels, the access point first assigns the active stations and then distributes the rest of the stations.
3.2. Scheduled Transmission (ST) Phase
All subchannels in the ST phase are assigned to only one station at a given time (i.e., OFDM operation), and stations are scheduled in time. The schedule and order of access to the medium are determined by the access point and enforced through broadcast messages indicating the schedule, or through explicit poll (TO assignment) messages. Transmission by stations is only allowed in response to such messages. This policy along with a carrier sense mechanism is used to coordinate contention free access to the medium during the ST phase. This mechanism utilizes the interframe spaces described earlier.
The first message in ST is sent by the AP. If this message is a schedule announcement, the stations will send their responses in the specified order with CIFS spacing. Acknowledgements are sent after MIFS following the data frame. This process continues until the end of the scheduled TOs with CIFS spacing separating the TOs. Since MIFS CIFS, the AP can reclaim the channel by transmitting after an MIFS, and change the schedule or end the ST phase.
If explicit poll messages are used instead of transmission schedules, the AP will send the first poll (TO assignment) message after CIFS MIFS. All message exchanges during a transmission opportunity use MIFS spacing. The use of time-based TO allows for multirate operation of the PHY while maintaining the temporal fairness (as discussed in other works ). The next TO is always initiated by the AP after a CIFS. This process continues until the end of the ST phase when the AP either indicates the end through an ST end frame or leaves the channel idle for at least RIFS CIFS (triggering a TR phase). Figure 2 demonstrates the timeline of the ST phase in two modes: using the explicit polling and using schedule announcement messages.
There are several options for enhancing the efficiency of the ST phase. One option is to piggyback the poll messages on the acknowledgement packets if the recipient of the station's packet is the access point (which is the case in most scenarios). Further enhancements can be achieved using block acknowledgment.
3.3. Quality of Service and Multimedia Support
To enable QoS and multimedia provisioning for the proposed MAC, we propose two classes of service: prioritized random access and scheduled guaranteed access. To provide priority services, we specify different limits for contention window sizes. The smaller the limits are, the higher the priority of access to the channel will be. In this case, it is also required that the access point schedules TOs for higher priority stations ahead of the others (using a simple priority scheduler).
The scheduled guaranteed access mechanism is usually used for traffic of persistent session (e.g., voice or video) type. This mode requires more complex scheduling schemes in the access point. The AP must also be aware of the traffic behavior of the requesting stations and their QoS requirements. When a traffic stream is set up between the AP and a station, the scheduler inside the AP is configured to send unsolicited TO assignments (e.g., polls) to the station. For this purpose any scheduling mechanism can be used. An example of such mechanisms is the controlled access phase scheduling mechanism , which was originally developed for 802.11e MAC but is applicable with modifications to the MAC protocol proposed here.
4. Analysis and Performance Evaluation
To analyze the performance of the proposed MAC, we first analytically model the MAC operation and derive the throughput in saturation mode (in which all stations always have data to send). We extend the analysis by simulation experiments for different traffic load conditions.
4.1. Analytical Modeling of the MAC in Saturation Mode
Knowing that each subchannel employs a CSMA/CA scheme with exponential random backoff, similar to the distributed coordination function (DCF) of 802.11, we can reuse the model developed for its backoff process . This model has been shown to accurately model the exponential backoff process in saturation mode [14–16].
The above probabilities are valid for our system and 802.11 DCF; however, their meanings are different in each MAC. For DCF, these probabilities describe the transition probabilities between successful collision and idle states. Whereas, in our system they only describe the probabilities that each of the backoff slots in a TR phase subchannel contains a transition to successful collision or idle states. For our MAC scheme, the value of in the above equations changes to the number of stations assigned to each subchannel.
where is the propagation delay, and are the PHY operational and basic rates, for example, = 54 Mbps and = 6 for IEEE the 802.11 a/g standard. Also, is the length of the Ack message and is the length of the PHY header.
The MAC protocol proposed in this article combines OFDMA with CSMA/CA mechanisms and significantly increases the performance and utilization efficiency of a WLAN in the MAC layer. Our results indicate that our protocol works best when the ratio of stations to channels is about 4 or 8. Below and above these ratios, performance tends to degrade. That may become a bottleneck if the AP can only offer a few channels; say 4. From practical perspective, one does not expect more than 20 to 30 stations to be associated with the same AP, making our scheme suitable for most practical WLANs.
We have selected OFDMA as the basis of our system due to its several advantages over other systems. Compared to CDMA systems, OFDMA can combat fading with less complexity. OFDMA can also achieve higher spectral efficiency. In comparison to TDMA-based systems, our system has a simpler random access scheme, that does not require synchronization, and it is suitable for a combination of data and multimedia traffic.
We presented the fundamental regulations and requirements of the proposed OFDMA-based MAC and developed a model for saturation throughput analysis. This model can be used for dynamically adjusting the number of subchannels and subchannel assignment to achieve optimal performance. Devising more complex analytical models and optimization algorithms for the OFDMA-based MAC is an interesting area of research. Another research subject that can be based on the proposed MAC is the design of scheduling algorithms for the contention free phase of the MAC operation. Such schedulers are required for QoS provisioning in the MAC layer.
This research was supported by a research grant from Qatar Foundation, under the National Priorities Research Program (NPRP) Grant 26-6-7-10.
- Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications ANSI/IEEE Std 802.11: 1999 (E) Part 11, ISO/IEC 8802-11, 1999Google Scholar
- Amendment to IEEE Standard 802.11e : Medium Access Control (MAC) Quality of Service (QoS) Enhancements. July 2005.Google Scholar
- Broadband Radio Access Networks (BRAN); HIPERLAN Type 2 Functional Specification; Data Link Control (DLC) Layer;Partl: Basic Data Transport Function In ETSI Report. ETSI, Nice, France; 2000.Google Scholar
- Bianchi G: Performance analysis of the IEEE 802.11 distributed coordination function. IEEE Journal on Selected Areas in Communications 2000, 18(3):535-547. 10.1109/49.840210View ArticleGoogle Scholar
- Chuang JC-I, Sollenberger N: Beyond 3G: wideband wireless data access based on OFDM and dynamic packet assignment. IEEE Communications Magazine 2000, 38(7):78-87. 10.1109/35.852035View ArticleGoogle Scholar
- Wong CY, Cheng RS, Letaief KB, Murch RD: Multiuser OFDM with adaptive subcarrier, bit, and power allocation. IEEE Journal on Selected Areas in Communications 1999, 17(10):1747-1758. 10.1109/49.793310View ArticleGoogle Scholar
- IEEE Standard for Local and Metropolitan Area Networks Part 16: Air Interface for Fixed Broadband Wireless Access Systems ANSI/IEEE Std 802.16-2004 (Revision of IEEE Std 802.16-2001)Google Scholar
- Jamalipour A, Wada T, Yamazato T: A tutorial on multiple access technologies for beyond 3G mobile networks. IEEE Communications Magazine 2005, 43(2):110-117.View ArticleGoogle Scholar
- Cuomo F, Baiocchi A, Cautelier R: A MAC protocol for a wireless LAN based on OFDM-CDMA. IEEE Communications Magazine 2000, 38(9):152-159. 10.1109/35.868155View ArticleGoogle Scholar
- Zhang Z, Liu Y: Multichannel Aloha data networks for personal communications services (PCS). Proceedings of IEEE Global Telecommunications Conference (GLOBECOM '92), December 1992, Orlando, Fla, USA 1: 21-25.Google Scholar
- Choi Y-J, Park S, Bahk S: Multichannel random access in OFDMA wireless networks. IEEE Journal on Selected Areas in Communications 2006, 24(3):603-613.View ArticleGoogle Scholar
- Wang X, Xiang W: An OFDM-TDMA/SA MAC protocol with QoS constraints for broadband wireless LANs. Wireless Networks 2006, 12(2):159-170. 10.1007/s11276-005-5263-1View ArticleGoogle Scholar
- Yaghoobi H: Scalable OFDMA Physical Layer in IEEE 802.16 WirelessMAN. Intel Technology Journal 2004, 8(3):201-212.Google Scholar
- Fallah YP: Per-session weighted fair scheduling for real time multimedia in multi-rate Wireless Local Area Networks, Ph.D. thesis. University of British Columbia, Vancouver, Canada; March 2007.Google Scholar
- Fallah YP, Alnuweiri HM: Modeling and performance evaluation of frame bursting in wireless LANs. In Proceedings of the International Wireless Communications and Mobile Computing Conference (IWCMC '06), July 2006, Vancouver, Canada. ACM; 869-874.Google Scholar
- Robinson JW, Randhawa TS: Saturation throughput analysis of IEEE 802.11e enhanced distributed coordination function. IEEE Journal on Selected Areas in Communications 2004, 22(5):917-928. 10.1109/JSAC.2004.826929View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.