2015 International Conference on Computing, Networking and Communications, Green Computing, Networking and Communications Symposium

Application based Energy Consumption Characterization of IEEE 802.11n/ac Access Points ¨ un Demir∗ , G¨unes¸ Karabulut Kurt∗ , Mehmet Karaca† Mehmet Ozg¨ ∗ Wireless

Communication Research Laboratory, Department of Electronics and Communication Engineering, Istanbul Technical University, Turkey † AirTies Wireless Networks, Istanbul, Turkey {[email protected], [email protected], [email protected]}

Abstract—Due to the rapid increase in the number of deployments of wireless local area networks (WLANs), the problem of energy efficiency has to be addressed for these networks. Until recent developments in hardware and software, the access points (APs) were kept switched on all the times even if there are no users that need connection (e.g. at night). Today, APs can support sleep mode which enables an AP to reduce energy consumption when there is no data transmission. However, the current implementation of sleep mode performs in a binary format where an AP either can be in sleep mode or transmits with full power. This binary format cannot provide both users’ satisfaction in a data communication and energy efficiency at the same time. In this paper, unlike binary energy model we design different energy modes which take into account not only the energy consumption but also quality of experience (QoE) at user side in terms of data rate. Index Terms—Energy consumption, Access Point, IEEE 802.11ac, IEEE 802.11n, Application requirement mode

I. I NTRODUCTION Today Internet and its applications have a great impact in our daily life. Inherently, people use Internet at almost every places such homes, offices, airports, etc., where wireless local area networks (WLANs) are very popular, and commonly used. As the demand of WLANs for high data rates rapidly increases especially after the huge increase in the number of internet users and applications, new WLAN standards are proposed regularly. Nowadays, IEEE 802.11n based Access Points (APs) are commonly used. However, in next couple of years, APs are designed according to IEEE 802.11ac WLAN standard that is expected to be used both indoor and outdoor places. The difference between these two standards is the achievable throughput on user side, so determination of standard is critical for Internet applications such as video conferencing, audio listening or sending e-mail. On the user side of these applications, customers experience includes customer satisfaction which is measured by the quality of experience (QoE). The field of green technology, targeting efficient usage of energy resources, is becoming more crucial day by day. Similarly, green systems in information and communication technologies (ICT) are becoming more critical as researchers This work was supported by TUBITAK under Grant 3130876.

978-1-4799-6959-3/15/$31.00 ©2015 IEEE

are constantly investigating new solutions. For example, several MAC protocols which aim to save energy are proposed. The solutions at physical (PHY) and application layers are developed to reduce energy consumption [1]. The sleep mode is very frequently investigated by researchers and it can be described as a mode that an AP reduces its energy consumption and senses the channel periodically when there is no transmission [2], [3]. Recently, many new APs were developed with sleep mode feature. However, using only these sleep modes offers binary mode levels, so it does not represent user behavior. Also using sleep mode reduces energy consumption only when there is no data transmission. In our paper, we extend the mode concept to other scenarios where there is not only sleep mode but also other energy modes developed for APs. We define a mode as a set of AP configurations which is determined according to users behavior, and targets to save energy. However, sleep modes are not compared in this research because we are only interested in new modes. Finding wider solutions which include end-user perspective, offering reducing energy consumption during data transmission and satisfying the minimum QoE are very important. With these objectives, different transmission modes addition to sleep mode can be described for different user behaviors. In this work, three modes (High, Moderate and Low performance) are defined for IEEE 802.11ac and two modes(High and Low performance) are defined for IEEE 802.11n. Firstly, in this paper, the tradeoffs between data rates and energy consumption are investigated through experimental studies. Then, our aim is to define the most energy efficient configuration at the AP for each user. Also, this configuration must support the minimum data rate requirement of each user by selecting the best PHY layer configuration. Hence, the energy consumption of AP and throughput are measured for different PHY configurations such as bandwidth (W ), transmit power (Pt ), transmit (Tx ) and receive (Rx ) antenna selection. After deciding energy efficient scenarios, they can be clustered in several modes. Then APs can dynamically select proper mode for the target application and also it should ensure both QoE and energy efficient transmission. Main contributions of this work are listed below:

520

2015 International Conference on Computing, Networking and Communications, Green Computing, Networking and Communications Symposium

•

•

In our experimental study, we measure energy consumption and throughput performance of IEEE 802.11n and IEEE 802.11ac based AP with several PHY layer configurations. We develop mode descriptions that are proposed according to end-user applications, and guarantee both energy efficiency and QoE.

A. Related Literature Many researchers investigate energy consumption models and energy reduction techniques in literature. Assuming only 802.11n standard, energy consumption measurements with different PHY configurations such as channel bandwidth, transmit power, data rates, antennas, multiple input multiple output (MIMO) streams, sleep and active modes are investigated in [4]. In [5], the authors measure the energy consumption in Joule/bit for cellular networks. In [6]–[11] the works focus on modeling energy consumption of an AP and in the common setup the iperf is used as a traffic generator to transmit UDP streams. [8] investigated the basic relationship between traffic and power consumption for IEEE 802.11g WLAN standard. Researchers are interested in the overall power consumption and the effect of different traffic patterns, data rates, packet sizes, security schemes, number of clients etc., on energy consumption in [11]. In [9], energy consumption of an AP is measured for different packet sizes and transmission rates at 5GHz industrial scientific and medical (ISM) band for IEEE 802.11ac and IEEE 802.11n. Also they investigated energy efficiency of emulations of YouTube and Skype flows. Moreover, there are various different proposals to reduce the overall energy consumption of wireless APs. These mainly target the adjustment of sleep modes, medium access control (MAC) level improvements or cross-layer parameter optimization techniques. In [12], authors propose a sleep mode by clustering access points and turning them on or off according to the usage patterns of client devices. In [13], energy efficiency and performance of WLANs are quantified and an energy efficient MAC protocol that dynamically adapts the interval of the announcement traffic indication message (ATIM) is proposed. Controlling the transmit power of APs in active states and associated power control algorithms are proposed in [14]–[17], demonstrating that a significant reduction of power consumption is possible when adapting the physical layer parameters according to the channel state information. Traffic scheduling idea along with the service requirements are also considered to reduce the overall consumed energy while addressing end user demands [18], [19]. Additionally, there are various patents and patent applications proposing methods to reduce energy consumption of wireless gateways. Adjustment of the sleep cycle by dividing the sleep/awake modes into one or more sleep sub-states is proposed in [20], that can partially activate the transmitter to increase power savings. QUALCOMM, in [21], a method to configure the APs to a low powered state for a pre-defined amount of time is proposed in order to extend the battery life.

Fig. 1. The measurement set-up. The server sends TCP packets to the client over the AP and power meter measures power consumption of the AP during transmission.

Also concentrating on user activity, QUALCOMM filed [22], to define AP rules to enter power-save mode. II. BACKGROUND I NFORMATION In this section, basic terms about IEEE 802.11 n/ac standards are given. Then, mode descriptions will be given along with the approach to reduce energy consumption of an AP. A. IEEE 802.11n and IEEE 802.11ac Overview In 2009, IEEE presented 802.11n wireless standard and this one offered faster data rates (up to 600 Mbps) than other 802.11 family members. Orthogonal frequency division multiplexing (OFDM) is used as the transmission technique. Multiple input multiple output (MIMO) was used first time in 802.11 standards and operating frequencies 2.4 GHz and 5 GHz are both available for transmission. Legacy IEEE 802.11 standards work only 20 MHz bandwidths, but the mandatory bandwidths of IEEE 802.11n are 20 and 40 MHz. At 2.4 GHz, both bandwidth are useful, but at 5 GHz only 40 MHz can be used with channel bonding. Moreover, this standard supports 64-quadrature amplitude modulation (QAM) with 5/6 coding rate [23], [24]. After four years, IEEE 802.11ac WLAN standard was introduced. The main goal of this standard is to reach Gbps data transmission rates [25]. OFDM is again used, similar to the legacy IEEE 802.11 family standards. This standard only available at 5 GHz and static or dynamic channel bonding is applicable. Multi-user MIMO (MU-MIMO) is critical to reach the target data rates. The standard also defines wider bandwidth usage; 20,40 and 80 MHz bandwidths are mandatory. 80 + 80 and 160 MHz bandwidths are kept optional for designers. In addition, maximum spatial stream number which is 4 in IEEE 802.11n case increased to 8. After all of these innovations, 7 Gbps data rate at PHY layer is possible [24], [26]. For testing AP, the highest possible data rates according to hardware is calculated 3.47 Gbps for IEEE 802.11ac standard [26].

521

2015 International Conference on Computing, Networking and Communications, Green Computing, Networking and Communications Symposium

Fig. 2. Energy consumption results for each configuration and various Pt values. It can be seen energy consumption levels represented by colors at colorbar on right side. On x-axis, labels are given in bandwidth : active Tx antenna : selected WLAN standard format. For example 20 : 2 : n refers that W = 20 MHz, 2 Tx antenna active and IEEE 802.11n standard configuration is selected.

B. Mode Description There are various internet applications chosen by internet users and requirements of these applications differ from each other. Many internet application requires different data rates for good QoE and generally APs offer higher data rates than required one, consequently, APs waste energy. For this reason, the main goal is describing modes to work APs energy efficiently and supply demanding data rates. As it is mentioned previously, determined AP configuration set according to application requirements is called mode. In Table I shows different internet applications and minimum required data rates for satisfactory QoE [27], [28].

Joule/MegaBits

0.4 0.3 0.2 0.1 0

6dBm 8dBm 10dBm 12dBm 14dBm Transmit Power

Fig. 3.

80:3:11ac 80:2:11ac 80:1:11ac 40:3:11ac 40:2:11ac 40:2:11n 40:1:11ac 40:1:11n 20:3:11ac 20:2:11ac 20:2:11n 20:1:11ac 20:1:11n

Joule/Bit results for each configuration and various Pt values.

TABLE I I NTERNET APPLICATIONS AND THEIR MINIMUM DATA RATES FOR SATISFACTORY Q O E AND ALSO POSSIBLE AP MODES . Internet Application M2M VoIP Hi-Fi Audio Streaming HDTV Video Streaming Sleep

Required Data Rate 8 Kbps 100 Kbps 5 Mbps 20 Mbps 50 Mbps

III. P OWER M EASUREMENT T ESTBED

Mode Name Very Low Very Low Low Moderate High Sleep

According to these data rates, 4 different transmission modes are needed. Also, AP should go into sleep cycle when there is no transmission. With sleep mode, total of 5 modes are described in Table I. As a result, PHY layer configuration consumed minimum energy and satisfied requiring data rate should be select for every mode. According to the [27], there is a bottleneck for house communication infrastructure, as a result indoor home capacity should be larger 5−10 times than data rates provided by access network connection. After that, required data rates for sufficient QoE and each application written on the Table I after they are multiplied by 7.

For measuring energy consumption of an AP supported IEEE 802.11n and 11ac standards, a testbed is designed and implemented. We inspired to construct this testbed from other related works [8], [11]. In this setup, two desktop PCs, one of them worked as a server, a laptop worked as a client, an AP and one power meter are used for all measurements as shown in Figure 1. AP has a wired ethernet connection to server and a wireless connection to the client. For observing the effects different PHY configuration on energy consumption, we changed Pt , bandwidth W and active Tx antennas for each IEEE 802.11 standards and fixed traffic. Iperf is selected as a traffic generator and transmission control protocol (TCP) packets sent from server to client are used. In IEEE 802.11n case, 2 transmit antennas are available and there are 3 antennas for IEEE 802.11ac case. In Table II, 3 Tx antenna combination for IEEE 802.11n are considered and there are 7 different Tx antenna configurations for IEEE 802.11ac measurements. Pt values are

522

2015 International Conference on Computing, Networking and Communications, Green Computing, Networking and Communications Symposium

400 350

Data Rate (Mbps)

300 250

6dBm 8dBm 10dBm 12dBm 14dBm

M2M HiFi Audio HD TV Video VoIP

200 150 100 50 0

20:1:n

20:1:ac

20:2:n

Fig. 4.

20:2:ac

20:3:ac

40:1:n

40:1:ac

40:2:ac

40:3:ac

80:1:ac

80:2:ac

80:3:ac

Data rates for each configuration and various Pt values. TABLE III IEEE 802.11 AC MODE PHY LAYER CONFIGURATIONS

TABLE II T EST C ONFIGURATIONS FOR IEEE 802.11 N AND 802.11 AC 802.11n Configurations Bandwidth Tx # of Tx Power Antenna 20 MHz 6 dBm 1 40 MHz 8 dBm 2 10 dBm 12 dBm 14 dBm

40:2:n

802.11ac Configurations Bandwidth Tx # of Tx Power Antenna 20 MHz 6 dBm 1 40 MHz 8 dBm 2 80 MHz 10 dBm 3 12 dBm 14 dBm

Mode Very Low Low Modarate High Sleep

changed between 6 and 14 dBm and difference between each transmit power value is 2 dBm, so there are a total of 5 distinct values for Pt . As aforementioned in Section in II-A, IEEE 802.11n standard supports 2 bandwidths (W = 20, 40 MHz) and W = 20, 40, 80 MHz bandwidths are possible for IEEE 802.11ac standard. To sum up, we measured power consumption of an AP for 30 different PHY configuration for IEEE 802.11n case and 105 for IEEE 802.11ac case. In this setup, Wattsup Pro is used as a power meter and it logs power consumption every second. At the same time, data rates were also recorded in every second, and the total measurement time is selected 50 seconds. The distance between AP and client is adjusted 1 meter. IV. A NALYSIS AND M EASUREMENT R ESULTS The measurements can be divided into 2 cases, first one is IEEE 802.11n measurements and others for IEEE 802.11ac. A total of 30 different PHY layer configurations are considered for first case and total of 105 scenarios are measured for the second one. The average results are considered. Also same tests are repeated for several times to ensure that results are stable. In Figure 2 energy consumption results are shown as thermal images. This figure indicates very clearly that using 2 transmitter antenna consumes less energy except some Pt = 6 dBm configurations. There are two important results of this figure. Firstly, increasing Pt causes more energy consumption. And

W 20 20 40 80

Tx 1 1 2 3

Pt 6 dBm 6 dBm 8 dBm 12dBm

Rb Requirement 1 Mbps 35 Mbps 140 Mbps 350 Mbps

secondly, W = 40 MHz bandwidths consumed much more energy than W = 20. Figure 3 shows Joule/Megabit values over all possible Pt , W , Tx antenna number for both IEEE 802.11n and 802.11ac. These values are found by dividing energy consumption values (in terms of Watt) to data rates (in terms of Megabit per second). When we interpret this figure, it is clear that there is a reduction for going to W = 80 MHz direction except W = 20 MHz 2 transmit antenna and IEEE 802.11n case. The application based modes we described can be seen clearly in Figure 4. Also configurations which can provide necessary QoE by supplying enough data rate can be select using this figure. According to this figure, if number of transmitter antenna or W increases, data rates also increase. In addition, data rates are not affected critically from increasing Pt at wider bandwidths. During mode selection, and we interested in both Figure 2 and 4 and looked for configurations satisfied data rate requirements and had low power. After that the mode configurations are decided for both of two standards. Selected mode configurations for IEEE 802.11ac are summarized in Table III. W values, numbers of Tx antenna and Pt levels are given with data rate requirTents and mode names. It can be seen that, actually there is no need to Very Low performance mode. Low performance mode can be used for machine-to-machine (M2M) communications, voice communications with Voice over Internet Protocol (VoIP) and listening high quality audio using internet. Moderate mode is designed for watching high definition TV (HDTV) broadcasts

523

2015 International Conference on Computing, Networking and Communications, Green Computing, Networking and Communications Symposium

TABLE IV IEEE 802.11 N MODE PHY LAYER CONFIGURATIONS Mode Low High Sleep

W 20 40

Tx 1 2

Pt 6 dBm 6 dBm

Rb Requirement 50 Mbps (Minimum) 100 Mbps (Maximum)

and operating other low data rate requirement applications. Moreover, most data rate need applications such as realtime video streaming and online gaming can be succeed running at High performance mode. IEEE 802.11n modes are divided in 2 as High and Low performance mode. For this AP, the highest data rate seen for IEEE 802.11n case is around 100 Mbps and the lowest one is near to 50 Mbps. Actually there is no huge difference between maximum and minimum data rate, in fact big differences occur at energy consumption results. For this reason, we decided to work AP in 2 modes and the mode configurations are given in IV. V. C ONCLUSION After measured energy consumption and data rates for many different PHY layer configurations of AP for both IEEE 802.11ac and 802.11n standards, we described internet application modes based on several common used internet applications. These modes satisfy necessary QoE levels for each users according to selected application based end-user behavior. Other feature of these modes is offering energy efficient management. As a future work these items can be considered: • Designing a transition model for moving between these modes • Changing AP configurations dynamically according to internet application • Reducing overall power consumption of AP according to user behavior using 2 previous items. R EFERENCES [1] T. Simunic, “Power saving techniques for wireless lans,” in Proceedings of the Conference on Design, Automation and Test in Europe - Volume 3, ser. DATE ’05. Washington, DC, USA: IEEE Computer Society, 2005, pp. 96–97. [2] S. Tang, H. Yomo, Y. Kondo, and S. Obana, “Exploiting burst transmission and partial correlation for reliable wake-up signaling in radioon-demand wlans,” in Communications (ICC), 2012 IEEE International Conference on, June 2012, pp. 4954–4959. [3] H. Han, Y. Liu, G. Shen, Y. Zhang, Q. Li, and C. Tan, “Design, realization, and evaluation of dozyap for power-efficient wi-fi tethering,” Networking, IEEE/ACM Transactions on, vol. 22, no. 5, pp. 1672–1685, Oct 2014. [4] D. Halperin, B. Greenstein, A. Sheth, and D. Wetherall, “Demystifying 802.11n power consumption,” in Proceedings of the 2010 International Conference on Power Aware Computing and Systems, ser. HotPower’10. Berkeley, CA, USA: USENIX Association, 2010. [5] C. Han, T. Harrold, S. Armour, I. Krikidis, S. Videv, P. M. Grant, H. Haas, J. Thompson, I. Ku, C.-X. Wang, T. A. Le, M. Nakhai, J. Zhang, and L. Hanzo, “Green radio: radio techniques to enable energyefficient wireless networks,” Communications Magazine, IEEE, vol. 49, no. 6, pp. 46–54, June 2011.

[6] S. Keranidis, G. Kazdaridis, V. Passas, T. Korakis, I. Koutsopoulos, and L. Tassiulas, “Nitos energy monitoring framework: Real time power monitoring in experimental wireless network deployments,” SIGMOBILE Mob. Comput. Commun. Rev., vol. 18, Feb. 2014. [7] L. Wang and J. Manner, “Energy consumption analysis of WLAN, 2G and 3G interfaces,” in GreenCom, 2010 IEEE/CPSCom, Dec 2010. [8] K. Gomez, R. Riggio, T. Rasheed, and F. Granelli, “Analysing the energy consumption behaviour of WiFi networks,” in Online Conference on Green Communications (GreenCom), 2011 IEEE, Sept 2011. [9] M. Tauber, S. Bhatti, and Y. Yu, “Application level energy and performance measurements in a wireless LAN,” in GreenCom, 2011 IEEE/ACM International Conference on, Aug 2011, pp. 100–109. [10] M. Tauber and S. Bhatti, “The effect of the 802.11 power save mechanism (psm) on energy efficiency and performance during system activity,” in Green Computing and Communications (GreenCom), 2012 IEEE International Conference on, Nov 2012, pp. 573–580. [11] G. Palem and S. Tozlu, “On energy consumption of Wi-Fi access points,” in Consumer Communications and Networking Conference (CCNC), 2012 IEEE, Jan 2012, pp. 434–438. [12] A. Jardosh, G. Iannaccone, K. Papagiannaki, and B. Vinnakota, “Towards an energy-star wlan infrastructure,” in Mobile Computing Systems and Applications. HotMobile Eighth IEEE Workshop on, March 2007. [13] S.-L. Wu and P.-C. Tseng, “An energy efficient mac protocol for ieee 802.11 wlans,” in Communication Networks and Services Research, 2004. Proceedings. Second Annual Conference on, May 2004. [14] Z. Niu, Y. Wu, J. Gong, and Z. Yang, “Cell zooming for cost-efficient green cellular networks,” Communications Magazine, IEEE, vol. 48, no. 11, pp. 74–79, November 2010. [15] Y. Chen, S. Zhang, S. Xu, and G. Li, “Fundamental trade-offs on green wireless networks,” Communications Magazine, IEEE, vol. 49, no. 6, pp. 30–37, June 2011. [16] H. Bogucka and A. Conti, “Degrees of freedom for energy savings in practical adaptive wireless systems,” Communications Magazine, IEEE, vol. 49, no. 6, pp. 38–45, June 2011. [17] M. Yanovsky, A. Gorbenko, and V. Kharchenko, “Adaptive wifi systems: Principles of design and application support,” in Digital Technologies (DT), 2013 International Conference on, May 2013, pp. 203–206. [18] K.-W. Chin, “A green scheduler for enterprise wlans,” in Australasian Telecommunication Networks and Applications Conference (ATNAC), 2011, Nov 2011, pp. 1–3. [19] D. Boscovic, F. Vakil, S. Dautovic, and M. Tosic, “Pervasive wireless cdn for greening video streaming to mobile devices,” in MIPRO, 2011 Proceedings of the 34th International Convention, May 2011, pp. 629– 636. [20] S. Thandaveswaran, “Access point power save mechanism for wireless communication systems,” Dec. 17 2013, uS Patent 8,611,268. [21] B. Sampathkumar, “Dynamic energy saving mechanism for access points,” Sep. 24 2013, uS Patent 8,542,620. [22] S. Manickam and V. Pitchaimani, “System and method for access point power save,” Jun. 13 2013, uS Patent App. 13/313,924. [23] B. Corporation, “802.11n: Next-generation wireless LAN technology,” Broadcom Corporation, Tech. Rep., April 2006. [Online]. Available: http://www.broadcom.com/collateral/wp/802˙11n-WP100-R.pdf [24] C. S. Inc., “802.11ac: The fifth generation of Wi-Fi,” Cisco Systems Inc., Tech. Rep., January 2014. [Online]. Available: http://www.cisco.com/c/en/us/products/collateral/wireless/aironet3600-series/white˙paper˙c11-713103.html [25] “IEEE standard for information technology– telecommunications and information exchange between systems local and metropolitan area networks– specific requirements–Part 11,” IEEE Std 802.11ac-2013, pp. 1–425, Dec 2013. [26] O. Bejarano, E. W. Knightly, and M. Park, “IEEE 802.11ac: from channelization to multi-user MIMO.” IEEE Communications Magazine, vol. 51, no. 10, pp. 84–90, 2013. [27] R. Ferrus, O. Sallent, J. Perez-Romero, and R. Agusti, “A solution framework to provide management services for wireless communications in the digital home,” Communications Magazine, IEEE, vol. 50, no. 11, pp. 132–141, November 2012. [28] J.-P. Javaudin, M. Bellec, P. Jaffre, O. Hoffmann, A. Foglar, and O. Isson, “Inter-mac concept for gigabit home networks,” in Personal, Indoor and Mobile Radio Communications, 2009 IEEE 20th International Symposium on, Sept 2009, pp. 1–5.

524