Enhancement of Wide Bandwidth Operation in IEEE ac Networks

Enhancement of Wide Bandwidth Operation in IEEE 82.11ac Networks Seongho Byeon, Changmok Yang, Okhwan Lee, Kangjin Yoon and Sunghyun Choi Department of ECE and INMC, Seoul National University, Seoul, Korea {shbyeon, cmyang, ohlee, kjyoon}@mwnl.snu.ac.kr, schoi@snu.ac.kr Abstract IEEE 82.11 has evolved from 82.11a/b/g/n to 82.11ac in order to meet ever-increasing high throughput demand. The newest standard supports various channel widths up to 16 MHz by bonding multiple 2 MHz channels. In order to efficiently utilize multiple channel widths, 82.11ac defines two operations, namely, dynamic channel access (DCA) and dynamic bandwidth operation (DBO). In this paper, we reveal that the use of DCA improves channel utilization significantly, and DBO not only partly overcomes secondary channel hidden interference problems but also achieves better channel utilization. However, when a transmitter attempts to send an enhanced RTS frame before data transmission as part of DBO, the station has no way to know how much bandwidth will be used, thus leading to malfunction of virtual carrier sensing at neighboring stations. On the other hand, the use of enhanced RTS/CTS without hidden traffic wastes airtime significantly. To address these problems, we first define how to calculate an appropriate value of duration field for enhanced RTS/CTS, and then develop an algorithm which adaptively enables/disables DBO considering the (secondary) hidden interference. Through ns-3 simulations, we demonstrate that the proposed scheme achieves up to 2x higher throughput compared to the baseline of 82.11ac. I. INTRODUCTION Over the last few years, wireless local area network (WLAN) technology has become an essential and indispensable part of our everyday life. Especially, Wi-Fi based on IEEE 82.11 WLAN standard has been one of the most successful commercial wireless technologies, supporting ever increasing demand of users for various types of services [1]. In this trend, IEEE 82.11 has evolved from 82.11a/b/g/n to 82.11ac to meet much-needed high throughput [2]. The achievable throughput of 82.11ac is at least 5 Mb/s for a single user, and at least 1 Gb/s in the case of multiple users. The newest standard supports the bandwidth of 2, 4, and 8 MHz as a mandatory feature, and optionally provides 16 MHz channel including non-contiguous 8+8 MHz [2, 3]. A wide bandwidth channel is divided into a primary channel and one or more secondary channels. Basically, only the primary channel follows the basic distributed coordination function (DCF) rules for contention, where the channel needs to be idle for DCF inter-frame space (DIFS) plus the backoff counter to reach zero. On the other hand, the secondary channels need to be idle during an interval of point coordination function (PCF) inter-frame space (PIFS) immediately preceding the expiration of the backoff counter of the primary channel. In case that any of the secondary channels is busy, then a station either restarts the channel Fig. 1. Channel allocation map in 5 GHz band. access (in the case of static channel access (SCA)) or transmits a frame including the primary channel and adjacent PIFS-idle secondary channels (in the case of dynamic channel access (DCA)). Besides, for the better bandwidth coordination, the standard provides dynamic bandwidth operation (DBO) which allows both transmitter and receiver to share the available bandwidth information using enhanced request-to-send (RTS) and clear-to-send (CTS) [2, 4]. This paper provides important lessons for employing 5 GHz channelization with a full investigation of SCA/DCA and DBO. While most of today s 82.11ac chipsets do not support DCA and DBO [5], we find that DCA improves channel utilization at a transmitter, and DBO not only overcomes the (secondary channel) hidden interference problem using enhanced RTS/CTS, but also achieves better channel utilization. Therefore, we claim that both DCA and DBO should be implemented in 82.11ac chipsets in order to efficiently utilize multiple channel widths. Even though DBO is likely to alleviate the (secondary channel) hidden interference problem, however, it has two critical problems. First, when a transmitter attempts to send an enhanced RTS frame to a receiver, the station has no way to know how much bandwidth will be used for the forthcoming data transmission. Therefore, 82.11ac standard does not define how to set the duration field value in enhanced RTS/CTS frames, and hence, virtual carrier sensing using enhanced RTS/CTS does not work properly. Moreover, enhanced RTS/CTS themselves could be protocol overheads since RTS/CTS handshake wastes the transmission airtime significantly and incurs excessive network allocation vector (NAV) settings. To address these problems, we first define how to calculate an appropriate value of duration field for enhanced RTS/CTS in order to conduct precise virtual carrier sensing, and then develop an algorithm which adaptively enables/disables DBO considering the (secondary) hidden interference. Through ns-3 simulations, we demonstrate that the proposed scheme achieves up to 2x higher throughput compared to the baseline of 82.11ac. Since the publication of IEEE 82.11ac standard [2], several

Channel# 36 4 44 48 52 56 6 64 Secondary 2MHz Fig. 2. Primary 2MHz Secondary 4 MHz Secondary 8 MHz Primary and secondary channel compositions. papers have provided an overview of 82.11ac amendment. In [3] and [6], the authors have introduced the core 82.11ac technologies, especially, wide bandwidth operation and multiuser multiple input multiple output (MU-MIMO). There have been studies on 82.11ac wide bandwidth operation in the literature [4, 7]. The work in [7] provides an insight of the necessity of dynamic bandwidth switching, called DCA, through simulation results. In [4], Gong et al. propose a MAC protection scheme, called DBO, that combats hidden nodes on secondary channels. As an extension, we consider both DCA and DBO simultaneously, and investigate the overall performance through ns-3 simulation. Recently, several measurement studies have been conducted for 82.11ac wide bandwidth operation [8, 9]. The study in [8] shows measurement results of wide bandwidth operation in an indoor environment, and the work in [9] focuses on the impact of wide bandwidth operation on both energy efficiency and secondary hidden interferences using 82.11ac testbed. Additionally, Jang et al. have proposed a channel allocation algorithm including the primary channel selection in 82.11ac environments for a given wider bandwidth [1]. This work also addresses hidden channel problem due to the difference of CCA sensitivity level and the difference of per-2 MHz spectral densities. However, none of these studies deals with DCA and DBO, since most commercial 82.11ac chipsets do not support DCA and DBO [5]. The rest of the paper is organized as follows. Section II introduces the background for 82.11ac wide bandwidth operation. In Section III, we provide in-depth performance analysis of SCA/DCA and DBO, and then identify the abovementioned problems. The detailed design of our scheme is presented in Section IV, and simulation results are presented in Section V. Finally, Section VI concludes the paper. II. BACKGROUND We dedicate this section to describing the channelization of IEEE 82.11ac for wide bandwidth operation as well as contention mechanism including both SCA and DCA. Furthermore, we also describe DBO feature which is an application of very high throughput (VHT) RTS/CTS, called enhanced RTS/CTS [2]. 1 A. Channelization IEEE 82.11ac mandates the operation of 2, 4, and 8 MHz channels, and optionally provides 16 MHz channel including non-contiguous 8+8 MHz channel. Fig. 1 shows channel allocation map in 5 GHz band. 4 MHz channel consists of two adjacent 2 MHz channels, and two adjacent 4 MHz channels form 8 MHz channel. Additionally, 1 The term very high throughput (VHT) stands for 82.11ac amendment. 16 MHz channel can be merged by two adjacent or separated 8 MHz channels [3, 7]. The wide bandwidth channel such as 4, 8, and 16 MHz consists of the primary 2 MHz channel and one or more secondary channels. For example, if channel 4 is selected as the primary channel, channel 36 corresponds to the secondary 2 MHz and channel 44 and 48 accord with the secondary 4 MHz channel as shown in Fig. 2. In order to guarantee the backward compatibility with legacy 82.11 specifications and the coexistence with other 82.11ac devices, only the primary 2 MHz channel performs a full clear channel assessment (CCA) including both physical carrier sensing (preamble detection) and virtual carrier sensing. Additionally, the primary channel and secondary channels have different rules for setting the network allocation vector (NAV). Even if the reception occurred only on the secondary channels, a station is not required to set the NAV. Moreover, the CCA sensitivity of the primary channel and secondary channels are different as well. CCA in 82.11ac is composed of signal detection (SD) and energy detection (ED). When an ongoing transmission of valid 82.11 signal occupies the primary channel, SD thresholds are 82 dbm for 2 MHz signals, and it becomes 3 db higher for wider bandwidth gradually. Otherwise, if an ongoing transmission of valid 82.11 signal does not occupy the primary 2 MHz channel, SD threshold for the secondary channels becomes 72 dbm which is 1 db lower than SD threshold of the primary channel. Table I summarizes SD thresholds used in 82.11ac. Note that when an ongoing transmission occupies the wider channel bandwidth than an operating bandwidth, SD threshold is determined based on the receiver s operating bandwidth, since a preamble of 82.11 signal has a 2 MHz duplicate structure. We mark red on the relevant values in Table I. Meanwhile, ED is performed when a signal is not decodable, so that ED does not require the assessment of whole operating bandwidth. Therefore, ED is executed on the primary 2 MHz channel and secondary 2, 4, and 8 MHz channels independently, using ED threshold of 62, 59, and 56 dbm, respectively [9]. TABLE I SIGNAL DETECTION THRESHOLDS (dbm). Receiving Operating 2 MHz 4 MHz 8 MHz 16 MHz Pri. Non-pri. Pri. Non-pri. Pri. Non-pri. Pri. Non-pri. 2 MHz 82 / 82 / 82 / 82 / 4 MHz 82 72 79 / 79 / 79 / 8 MHz 82 72 79 72 76 / 76 / 16 MHz 82 72 79 72 76 69 73 / B. Contention Mechanism One of the important aspects of 82.11ac is the contention and access mechanism for wide bandwidth operation. Basically, IEEE 82.11 defines DCF contention mechanism, where the channel needs to be idle for DIFS plus the backoff counter which is randomly selected within contention window to reach zero. The primary 2 MHz channel follows this DCF rule, however, all secondary channels have a different criterion where the channels need to be idle during an interval of PIFS immediately preceding the expiration of the primary channel s

(a) Static channel access (SCA). (b) Dynamic channel access (DCA). Fig. 3. Two types of channel accesses for 82.11ac wide bandwidth operation. Algorithm 1 Wide bandwidth operation in 82.11ac 1: Transmitter needs to wait until the primary channel is idle during DIFS plus backoff counter. At the moment the backoff counter expires, if the secondary channels are all idle, sends the enhanced RTS frame in the format of duplicate 82.11 legacy through BW op, otherwise uses BW tx which includes the primary 2 MHz channel and adjacent PIFS-idle secondary channels. 2: Receiver sends back the enhanced CTS through BW rx including the primary channel and PIFS-idle secondary channels within the bandwidth specified in the enhanced RTS. 3: Transmitter sends BW rx PPDU, if successfully receives the enhanced CTS, otherwise goes back to the first step. 4: Receiver sends duplicate block acknowledgement (block- Ack), when successfully receives BW rx PPDU. backoff counter. Then, in order for a station to transmit wide bandwidth channel PHY convergence procedure protocol data unit (PPDU), the station transmits BW op PPDU if the primary 2 MHz and corresponding secondary channels are all idle, where BW op {2, 4, 8, 16 MHz}. On the other hand, if any one of the secondary channels in BW op is busy, the station should perform exactly one of the followings: (i) Static channel access (SCA): Restart the channel access attempt by invoking the backoff procedure, where the station selects a new random number using the current value of contention window. (ii) Dynamic channel access (DCA): Transmit BW tx PPDU including the primary 2 MHz and the contiguous idle secondary channels, where BW tx {2, 4, 8, 16 MHz} and BW tx BW op. Fig. 3 illustrates examples of static/dynamic channel access [3, 7]. DCA allows stations to achieve more efficient channel utilization than SCA. C. Dynamic Bandwidth Operation (DBO) Enhanced RTS/CTS frames contain bandwidth information for forthcoming PPDU transmission, using the first 7 bits of scrambling sequence [4]. As an application of it, the standard defines dynamic bandwidth operation (DBO) as an optional feature, which allows the transmitter and receiver to negotiate a potentially reduced bandwidth by informing their available bandwidth through RTS/CTS exchange. The transmitter first sends an enhanced RTS frame in the format of duplicate 82.11 legacy through wide bandwidth channel. Each of duplicate RTS frame contains the transmit bandwidth, BW tx, therefore, the receiver is aware of the transmitter s desired bandwidth for forthcoming data PPDU. After short inter-frame space (SIFS), the receiver sends back an enhanced CTS frame in a duplicate 82.11 legacy format only on the idle channels within BW tx. The idleness of each 2 MHz channel follows the rule as explained in Section II-B. The bandwidth carried by enhanced CTS, BW rx, could be reduced in comparison with BW tx, the bandwidth initially selected by the transmitter. Finally, the transmitter sends data PPDU using BW rx. Note that DBO not only alleviates the secondary channel hidden interference problem, 2 but also improves channel utilization by sharing the available bandwidth each other. Basically, it is well known that RTS/CTS handshake partly overcomes the hidden interference problem by inducing NAV setting to other 82.11 stations. Specifically, stations who receive RTS/CTS frames set the NAV as much as the time specified in the duration field of medium access control (MAC) header. The problem of DBO is that 82.11ac standard does not define how to set a duration field value in enhanced RTS/CTS frames. When a transmitter sends an enhanced RTS frame to a receiver, it does not know how much bandwidth will be used for data PPDU. Therefore, the accurate time carried by enhanced RTS cannot be determined in this step, since the transmission time of a PPDU varies according to different bandwidth values. Naturally, enhanced CTS also has no way to notice its duration value due to the absence of a reference. Consequently, the virtual carrier sensing using enhanced RTS/CTS might not work properly in 82.11ac. Furthermore, it is well known that RTS/CTS frames are overhead, since RTS/CTS handshake significantly wastes the transmission airtime and incurs excessive NAV settings. We will address these limitations in Section III. To sum up, Algorithm 1 describes the entire transmission and reception process of 82.11ac wide bandwidth operation on the operating bandwidth of BW op, assuming that both DCA and DBO are enabled. III. PROBLEM STATEMENT In this section, we first investigate the performances of SCA/DCA and DBO using network simulator (ns-3) [11]. We implement wide bandwidth operation of IEEE 82.11ac in ns-3, including DCA and DBO. For each set of simulation 2 The secondary channel hidden interference means a situation where a hidden interference disturbs the reception of PPDU only on some parts of the secondary channels.

Throughput (Mb/s) 4 35 3 25 2 15 5 Target (82.11ac) throughput Hidden (82.11n) throughput Sum throughput 5 1 15 2 25 3 35 4 45 5 Distance between APs (m) (a) Throughput results using the baseline of 82.11ac (SCA). Throughput (Mb/s) 4 35 3 25 2 15 5 SCA DCA DCA w/ DBO DCA w/ RTS/CTS Proposed scheme 5 1 15 2 25 3 35 4 45 5 Distance between APs (m) (b) Sum throughput results. Target BSS, 82.11ac Primary36,MCS7 Fig. 4. BW8MHz 5m Fig. 5. Distance between APs Throughput results for the case of a single station in each BSS. Hidden BSS, 82.11n Primary44,MCS7 BW4MHz 5m Simulation topology: a single station in each BSS. scenarios, we average the results of 2 simulation runs where each runs for 5 seconds simulation time. We also adopts the pathloss model with pathloss exponent of 3.5 for typical indoor environment, and the Rayleigh fading channel model with the Doppler velocity of.1 m/s. Fig. 4 illustrates a simulation topology configured to deploy two 82.11n/ac basic service sets (BSSs) with a single access point (AP) and one station, respectively. The 82.11ac BSS, called target BSS, uses channel 36 as the primary 2 MHz channel and its operating bandwidth is set to 8 MHz. On the other hand, the primary 2 MHz channel of 82.11n BSS, called hidden BSS, is channel 44 and its operating bandwidth is fixed to 4 MHz. Therefore, 82.11n signal occupies the secondary 4 MHz of 82.11ac BSS. We generate saturated UDP downlink traffic from each AP to its associated station, where the distance between AP and corresponding station is fixed to 5 m. The length of each MAC protocol data unit (MPDU) is 1,534 bytes including MAC header, and MPDUs are aggregated into an Aggregate MPDU (A-MPDU). Fig. 5(a) shows throughput results of each BSS using the baseline of 82.11ac (SCA) as the distance between APs increases. When the distance is shorter than 1 m, both APs can carrier-sense each other, where 82.11ac AP uses CCA sensitivity of 72 dbm for the secondary 4 MHz channel, and 82.11n AP adopts 79 dbm as explained in Section II-A. As the distance increases beyond 1 m, 82.11ac AP becomes unable to detect the signals from 82.11n gradually. Therefore, 82.11ac AP aggressively attempts to access the wireless medium even though 82.11n AP is transmitting a packet, which results in collision errors [7, 1]. This phenomenon is getting worse until the distance reaches to 19 m. If the distance becomes larger than 19 m, 82.11n AP also cannot detect the signals from 82.11ac. Interestingly, from this point, 82.11ac signal does not interfere with the reception performance of 82.11n station so that 82.11n BSS achieves the maximum throughput independently. On the other hand, 82.11ac station will receive packets without the impact of the (secondary channel) hidden interference when the distance becomes at least 35 m long. 3 We next analyze the performance of DCA and DBO on the same environments. Fig. 5(b) demonstrates sum throughput results of each scheme. DCA permits 82.11ac AP only to access non-overlapped 4 MHz bandwidth (channel 36 and 4), when 82.11n AP transmits a packet on its 4 MHz bandwidth. Consequently, it shows up to 26% higher throughput by achieving better channel utilization, when both APs can carrier-sense each other. Using DBO, throughput does not decrease in the range from 1 m to 14 m, since the transmitter and receiver can share their available bandwidth. Moreover, as long as 82.11ac station can notice the interference signals (above CCA threshold of the secondary channels), it always achieves up to 9% higher throughput than the baseline of 82.11ac. As a consequence, the use of DCA and DBO always provides the better performance. Accordingly, we insist that both of them should be utilized under any circumstances. For a fair comparison, we also investigate the performance of enhanced RTS/CTS without DBO. That is, unless the whole channel specified in enhanced RTS is idle, the receiver does not send back enhanced CTS to the transmitter. Therefore, the value of duration field in enhanced RTS/CTS could be determined in advance, since the transmitter can be aware of the bandwidth which will be used for forthcoming PPDU transmission, which is different from DBO. Since some of 82.11ac packets are protected by virtual carrier sensing, 4 it achieves up to 74% higher throughput when there exists hidden interference that cannot be detected even at the receiver as shown in Fig. 5(b). However, without hidden interference, not only enhanced RTS/CTS waste the airtime, but also excessive NAV setting reduces frequency reuse. As a result, throughput decreases by up to 12% at 3 m location. Based on the above observations, the duration value of enhanced RTS/CTS for DBO should be properly determined, since enhanced RTS/CTS used in DBO cannot configure its proper duration value as explained in Section II-C, which causes to lose the original purpose of RTS/CTS exchange. Furthermore, DBO should be enabled only when hidden interference exists. 3 Power of 8 MHz signal per 2 MHz is 3 db weaker than that of 4 MHz signal. Therefore, the interference range of 4 MHz signal is wider. 4 82.11n traffic cannot be protected by NAV setting since 82.11ac BSS cannot receive RTS/CTS from 82.11n in this channel allocation.

Tx Rx Others DIFS +Backoff Enhanced RTS[8] Fig. 6. SIFS T cts SIFS Enhanced CTS[4] NAV(RTS) Data[4] T d NAV(CTS) Effect of duration value configuration. SIFS T back blockack [4] IV. ENHANCING WIDE BANDWIDTH OPERATION In this section, we define how to calculate an appropriate value of duration field for enhanced RTS/CTS, and then develop an algorithm which adaptively enables and disables DBO depending on the intensity of hidden interference. A. Duration Value of Enhanced RTS/CTS DBO allows the transmitter and receiver to negotiate a potentially reduced bandwidth by informing their available bandwidth through enhanced RTS/CTS exchange. When the transmitter sends an enhanced RTS, it does not know how much bandwidth will be used for its data transmission so that the duration value of enhanced RTS cannot be determined in this step. Here, we propose a simple protocol to determine the value of duration field in enhanced RTS/CTS considering the bandwidth negotiation. The transmitter calculates its duration value, D rts, based on the amount of bandwidth through which enhanced RTS will be transmitted: D rts = 3SIF S + T cts + T d + T back, where T cts, T d, and T back represent the transmission airtime for CTS, data frame, and blockack, respectively. After receiving enhanced RTS, the receiver recalculates a duration value of enhanced CTS, D cts, according to an idleness of secondary channels: D cts = 2SIF S + min(αt d, ap P DUMaxT ime) + T back, where α is a ratio of the bandwidth specified in enhanced RTS to the bandwidth of enhanced CTS, and ap P DUMaxT ime indicates the maximum PPDU transmission time defined in the standard [2]. Fig. 6 represents an effect of the proposed protocol using DBO. Since the bandwidth is reduced by half, T d becomes double so that data transmission can be protected for longer time through enhanced CTS. B. Adaptive Use of DBO The proposed algorithm calculates the collision loss probability, P col, by utilizing A-MPDU transmission results from blockack. Especially, hidden interference leads to bursty MPDU losses and many retransmissions [12]. Therefore, if a transmitter does not receive blockack or A-MPDU experiences bursty MPDU errors, the proposed algorithm regards them as collision losses. To avoid using stale information, we adopt a time window, T wnd. Based on the statistics collected during T wnd, the proposed algorithm calculates P col and compares it with P overhead which represents a ratio of throughput loss due to the overhead of RTS/CTS handshake Algorithm 2 Adaptive use of DBO Require: blockack bitmap vector 1: if DBO disabled 2: if blockack missing or consecutive MPDU losses 3: Update collision errors 4: if first instantaneous P col 9% 5: Turn on DBO for the current window 6: if T wnd expires 7: if DBO enabled then 8: Turn off DBO 9: else if P col P overhead then 1: Turn on DBO 11: Restart T wnd and reset all parameters Throughput (Mb/s) 5 4 3 2 Fig. 7. SCA DCA DCA w/ DBO DCA w/ RTS/CTS DCA w/ DBO (duration) DCA w/ DBO (on/off) Proposed scheme 5 14 2 36 Distance between APs (m) Throughput results for the case of a single station in each BSS. for a given bandwidth and PHY rate. If P col P overhead, DBO will be enabled for the next time window. As soon as this time window expires, if the collision probability for the first A- MPDU (first instantaneous P col ) becomes larger than 9% due to severe hidden interferences, the proposed algorithm enables DBO for the current window once again. Algorithm 2 shows the entire process of adaptive use of DBO. V. PERFORMANCE EVALUATION We evaluate the performance of the proposed scheme through ns-3 simulation. As comparison schemes, we choose the baseline of 82.11ac (SCA), DCA without DBO (DCA), DCA using DBO (DCA w/ DBO), and DCA using enhanced RTS/CTS without DBO (DCA w/ RTS/CTS). Additionally, we also compare the performance of the proposed protocol which calculates an appropriate value of duration field in enhanced RTS/CTS (DCA w/ DBO (duration)) and the proposed algorithm which adaptively enables/disables DBO (DCA w/ DBO (on/off)), respectively. The details of simulation environments are described in Section III. A. Single Station in Each BSS Initially, we deploy two 82.11n/ac BSSs with a single AP and one station as shown in Fig. 4. Fig. 7 shows sum throughput results of two BSSs for given AP distances. When the distance between APs is 5 m, both APs can carrier-sense each other, and hence, DCA achieves 26% higher throughput on average. At 14 m, 82.11ac AP cannot sense the signal from 82.11n AP, while 82.11ac station can detect it. Therefore,

2m 2BSSs 12m 8 MHz 82.11ac(primary 36) 4 MHz 82.11n (primary 44) 3BSSs 24m (a) Simulation topology: multiple stations in each BSS. Throughput (Mb/s) 3 25 2 15 5 SCA DCA DCA w/ DBO DCA w/ RTS/CTS DCA w/ DBO (duration) DCA w/ DBO (on/off) Proposed scheme 2 BSSs 3 BSSs (b) Sum throughput results for the case of multiple stations in each BSS. Fig. 8. Simulation topology and corresponding sum throughput results. the use of DBO always provides the higher throughput, increasing in channel utilization. Additionally, complete hidden relationships are established at 2 m, since received signal power is smaller than CCA threshold for both 82.11ac and 82.11n. Accordingly, NAV settings using appropriate duration value of enhanced RTS/CTS shows up to 12% higher throughput. Finally, when the distance between APs becomes 36 m, both signals do not interfere with each other, so each BSS becomes independent. Here, the use of enhanced RTS/CTS leads to an additional overhead as explained in Section III. Interestingly, our proposed scheme achieves almost the best throughput for any circumstances by accepting the benefits of other existing features. The throughput results at all AP distances are shown in Fig. 5(b). As expected, the proposed scheme follows the upper-most curves irrespective of the distance. B. Multiple Stations in Each BSS We also investigate the performance of the proposed scheme in multi-station environments. Four stations are randomly deployed within each BSS as shown in Fig. 8(a). We adopt an ideal rate adaptation algorithm that determines the best PHY rate without any overhead for a given time-varying channel quality. Fig. 8(b) shows the downlink sum throughput achieved by all BSSs. In 2 BSSs case, the distance between 82.11ac AP and 82.11n AP is set to 2 m where both APs cannot carrier-sense each other. In this scenario, if any one of stations is located near the hidden AP, DBO and/or NAV setting from enhanced RTS/CTS can partly overcome this hidden interference problem. Otherwise, if a station is located near the target AP, it might not be significantly affected from hidden interference due to relatively high signal power. Therefore, our proposed scheme, which adaptively uses DBO and sets a duration value of enhanced RTS/CTS appropriately, achieves 78.4% and 14.9% higher throughput than that of a baseline of 82.11ac (SCA) and DCA with DBO, respectively. This observation goes to the same with 3 BSSs case. Here, two 82.11n APs are deployed in 12 m leftward and 24 m rightward from 82.11ac AP, respectively. DBO maximize its benefit when the distance between APs is 12 m, and right NAV setting achieves the highest throughput at 24 m distance as explained in Section III. As a result, our proposed scheme shows 21.8% higher throughput than that of DCA with DBO. VI. CONCLUSION AND FUTURE WORK In this paper, we first investigate the performance of DCA and DBO in IEEE 82.11ac. While most of today s 82.11ac chipsets do not support these operations, we reveal that the use of DCA significantly improves channel utilization, and DBO not only partly overcomes secondary channel hidden interference problems but also achieves better channel utilization. Afterward, we define how to calculate an appropriate value of duration field in enhanced RTS/CTS, and then propose an algorithm that adaptively enables/disables DBO. Finally, we demonstrate that the proposed scheme outperforms up to 2x compared to the baseline 82.11ac. As a future work, we now study a joint rate and bandwidth adaptation considering both hidden interference level and time-varying channel quality. ACKNOWLEDGMENT This research was supported by the International Research & Development Program of the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT and Future Planning of Korea (NRF-214K1A3A7A37347), and the Brain Korea 21 Plus Project in 215. REFERENCES [1] Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, 213 218, http://www.cisco.com/. [2] IEEE 82.11ac, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications: Enhancements for Very High Throughput for Operation in Bands below 6 GHz, IEEE Std., Dec. 213. [3] O. Bejarano, E. W. Knightly, and M. Park, IEEE 82.11ac: From Channelization to Multi-User MIMO, IEEE Commun. Mag., vol. 51, no. 1, pp. 84 9, Oct. 213. [4] M. X. Gong et al., Channel Bounding and MAC Protection Mechanmsim for 82.11ac, in Proc. IEEE Globecom, Dec. 211. [5] List of 82.11ac Hardware, https://wikidevi.com/wiki/list of 82.11ac Hardware/. [6] E. Perahia and M. X. Gong, Gigabit Wireless LANs: An Overview of IEEE 82.11ac and 82.11ad, in Proc. ACM SIGMOBILE, Sep. 211. [7] M. Park, IEEE 82.11ac: Dynamic Bandwidth Channel Access, in Proc. IEEE ICC, Jun. 211. [8] M. D. Dianu, J. Riihijärvi, and M. Petrova, Measurement-Based Study of the performance IEEE 82.11ac in an Indoor Environment, in Proc. IEEE ICC, Jun. 214. [9] Y. Zeng, P. H. Pathak, and P. Mohapatra, A First Look at 82.11ac in Action: Energy Efficiency and Interference Characterization, in Proc. IFPF Networking, Jun. 214. [1] S. Jang and S. Bahk, A Channel Allocation Algorithm for Reducing the Channel Sensing/Reserving Asymmetry in 82.11ac Networks, IEEE Trans. Mobile Comput., vol. PP, no. 99, pp. 1 14, Jun. 214. [11] The Network Simulator 3 ns-3. http://www.nsnam.org/. [12] R. Anwar et al., Loss Differentiation: Moving onto High-Speed Wireless LANs, in Proc. IEEE INFOCOM, May 214.