Designing Energy-Efficient Wireless Access ... - Semantic Scholar

91 downloads 98363 Views 675KB Size Report
LTE and LTE-Advanced. LTE, standardized by the Third-Generation. Partnership Project (3GPP), is an emerging wireless access technology. Each new ver-.
Feature: Wireless Networks

Designing Energy-Efficient Wireless Access Networks: LTE and LTE-Advanced As large energy consumers, base stations need energy-efficient wireless access networks. This article compares the design of Long-Term Evolution (LTE) networks to energy-efficient LTE-Advanced networks. LTE-Advanced introduces three new functionalities — carrier aggregation, heterogeneous networks, and extended multiple-input, multiple-output (MIMO) support. The authors develop a power consumption model for LTE and LTE-Advanced macrocell and femtocell base stations, along with an energy efficiency measure. They show that LTE-Advanced’s carrier aggregation and MIMO improve networks’ energy efficiency up to 400 and 450 percent, respectively.

T

he last few years have seen a tremendous increase in the number of mobile users, with global mobile phone penetration increasing from 20 percent in 2003 to 67 percent in 2009.1 This growth has affected wireless access networks (WANs), which are already large energy consumers within the information communication technology (ICT) domain. Base stations account for up to 90 percent of a WAN’s power consumption.2 A thorough study of a base station’s power consumption can help us to develop guidelines for reducing these networks’ power consumption and thus reducing ICT’s ecological footprint. We investigate how energy-efficient Long-Term Evolution (LTE)-Advanced (release 10) access networks can be designed to improve energy efficiency compared to LTE (release 8/9) networks. SEPTEMBER/OCTOBER 2013

IC-17-05-Dery.indd 39

LTE-Advanced introduces three additional functionalities that influence energy efficiency:

Margot Deruyck, Wout Joseph, Bart Lannoo, Didier Colle, and Luc Martens Ghent University

• carrier aggregation, to increase the bit rate; • heterogeneous networks — that is, a combination of macrocell- and femtocell-based stations in one network; and • extended support for multiple-input, multiple-output (MIMO), allowing multiple antennas to send and receive signals. We develop a power consumption model for both LTE and LTE-Advanced macrocell and femtocell base stations, as well as a suitable energy-efficiency measure. Although some researchers have studied the energy efficiency of heterogeneous

1089-7801/13/$31.00 © 2013 IEEE

Published by the IEEE Computer Society

39

10/08/13 2:51 PM

Feature: Wireless Networks

LTE and LTE-Advanced networks,1,3,4 to the best of our knowledge, no one has compared the energy efficiency of LTE and LTE-Advanced or the effect of LTE-Advanced’s main functionalities on energy efficiency.

LTE and LTE-Advanced

LTE, standardized by the Third-Generation Partnership Project (3GPP), is an emerging wireless access technology. Each new version defined in the last few years has offered new features while preserving backward compatibility. The first two releases (release 8/9) are known as LTE.5 Release 10, also known as LTE-Advanced,5 supports even higher bit rates. Because LTE and LTE-Advanced are the same technology, LTE-Advanced is backward compatible with LTE. LTE and LTE-Advanced let devices adaptively change modulation (for translating the digital signal to an analog signal that can be transmitted wirelessly), coding rate (for detecting errors due to wireless transmission), and bandwidth to enhance channel quality. It supports bandwidths from roughly 1 to 20 MHz. This bandwidth can be further extended in LTE-Advanced by carrier aggregation, which lets the base station transmit multiple LTE carriers, each with a bandwidth of up to 20 MHz. That is, when no carrier aggregation is used, the user device will receive one carrier. When using carrier aggregation, it is possible to send not just one carrier but multiple carriers to the users, which results in a higher bit rate.6 A carrier, in this context called a component carrier, contains the data to be sent. Support for heterogeneous networks is further improved in LTE-Advanced.6 LTE-based heterogeneous networks are typically two-layered networks with macrocell (eNodeB) and femtocell (home-eNodeB) base stations. Although LTE already supports this functionality, LTEAdvanced improves the handling of interference between the different cells.6 Finally, LTE-Advanced enhances support for MIMO up to eight transmit antennas.6

Energy Efficiency in Wireless Access Networks

Determining which base station type is most energy efficient isn’t easy because we must consider different performance parameters (bandwidth, 40

IC-17-05-Dery.indd 40

www.computer.org/internet/

coverage, served users, and so on). Therefore, we define energy efficiency EE (∈ [0, ∞[) (in (km2 • Mbps)/W) as EE =

π ⋅ R2 ⋅ B ⋅ U ,(1) Pel

where Pel is the base station’s power consumption (in watts), R is the range (in km), B is the physical bit rate by the base station (in Mbps), and U is the number of served users. The higher the EE, the more energy efficient the base station is. Most measures in the literature consider bit rate, covered area, or number of served users. Others consider the power consumption per covered area4 or power consumption over capacity,7 which is meaningful when comparing the energy efficiency of base stations offering the same or similar capacity. Another measure, power consumption per user,8,9 is meaningful for wired technologies where users sharing the same modems are offered the same range and bit rate, but less meaningful for wireless technologies because range and bit rate can differ from base station to base station. Because our definition considers all these parameters, we investigate their aggregated influence on energy efficiency.

Power Consumption As Figure 1 shows, a macrocell base station consists of six power-consuming components2: • rectifier (100 W), which converts alternating current (AC) to direct current (DC); • digital signal processing (100 W), which converts the signal to a sequence of bits or symbols; • transceiver (100 W), which transmits and receives the signals; • power amplifier (156.3 W), which converts the DC input power into a significant RF signal; • air conditioning (225 W), which regulates the temperature in the base station cabin; and • backhaul (80 W), which oversees communication with the backhaul network (a microwave or fiber link). The sum of these components corresponds to the base station power consumption. However, in Figure 1, some components are used multiple times, and their power consumption should be multiplied by the number of occurrences. Which components to use and how many of them IEEE INTERNET COMPUTING

12/08/13 3:24 PM

Designing Energy-Efficient Wireless Access Networks: LTE and LTE-Advanced

depend on two factors: the number of sectors and the number of transmitting antennas (nT ). x We further divide the area covered by a base station, a cell, into several sectors, each covered by one antenna. For each sector, we need nT recx tifiers, nT digital signal processing components, x nT transceivers, and nT power amplifiers. We x x multiply these components’ power consumption by nsector. Unless otherwise mentioned, we assume one transmitting antenna per sector. Each base station component has a typical power consumption (Figure 1), which, except for the air conditioning and power amplifier, we assume to be constant throughout time. The power consumption of the latter depends on the antenna’s input power, PT .10 The higher the PT , the higher the x x power amplifier’s power consumption. For the macrocell base station, LTE and LTE-Advanced have a typical PT of 43 dBm. The power consumpx tion of the air conditioning component depends on the cabin’s internal and ambient temperature, which we assume to be constant at 25 °C. We assume that each component’s power consumption is constant throughout time. How­ ever, the power consumption of the digital s­ ignal processing, transceiver, and power amplifier components can fluctuate due to load ­variations on the base station, which represent the number of active users and the requirements of the used services in the base station cell. The higher the load, the higher the base station’s power consumption. To account for this fluctuation, we define the load factor as a value between 0 (no traffic) and 1 (maximum traffic). We multiply the power consumption of the digital signal processing, transceiver, and power amplifier components by this load factor, which varies between 0.93 and 1 during the day, to determine the base station’s power consumption at a certain time of day.2 Here, we assume a load factor of 1 (maximal power consumption of base station) and investigate the worst-case scenario. Accounting for this load factor, we obtained a power consumption of 1,672.6 W per base station (LTE and LTE-Advanced, nT = 1).2 x A femtocell base station is much smaller than a macrocell base station and is comparable to the base station of a Wi-Fi access point. As Figure 1 shows, the power-consuming components thus differ from those of a macrocell base station11,12: • microprocessor (3.2 W), which implements and manages the standardized radio SEPTEMBER/OCTOBER 2013

IC-17-05-Dery.indd 41

Macrocell

Femtocell nsector nTx

Transceiver (100 W)

nTx Transceiver (1.8 W)

nTx

Power amplifier (156.3 W)

Power amplifier (2.4 W)

Digital signal processing (100 W)

Mircoprocessor (3.2 W)

Rectifier (100 W)

FPGA (4.7 W) One sector

nTx

One sector

Air conditioning (225 W) Backhaul (80 W)

Figure 1. Block diagram of the power-consuming components of a macrocell and femtocell base station.10,11 ­ rotocol stack, the baseband processing, and p backhauling; • field-programmable gate array (FPGA; 4.7 W), which is responsible for features such as data encryption and hardware authentication; • transceiver (1.8 W); and • power amplifier (2.4 W). Figure 1 shows these components’ power consumption, which results in a femtocell base station power consumption of 12 W (nT = 1). x

Range of Base Station Types To determine a base station’s range, we calculate the maximum allowable path loss PLmax. Path loss is the ratio of the signal’s transmitted power to its received power.10 PLmax is thus the maximum allowable path loss to which a transmitted signal can be subjected while still being detectable at the receiver. Based on PLmax, we can determine the range using a propagation model that describes the relation between path loss and range. We must use a different propagation model for the two base station types because the circumstances for each type are different — for example, the femtocell base station is situated indoors, ­ 41

10/08/13 2:52 PM

Feature: Wireless Networks

Energy efficiency EE ((km2 ⋅Mbps)/W)

2.5

Users

2 1.5 1 0.5 0

1/2 QPSK

2/3 QPSK

Macro LTE =1×5 MHz Macro LTE Adv. CA 2×5 MHz Macro LTE Adv. CA 5×5 MHz

2/3 16-QAM

2/3 64-QAM

Femto LTE =1×5 MHz Femto LTE Adv. CA 2×5 MHz Femto LTE Adv. CA 5×5 MHz

Figure 2. Energy efficiency of LTE and LTE-Advanced in a 5-MHz channel. The figure compares the energy efficiency when aggregating no carriers (LTE) and when aggregating two or five carriers (LTE-Advanced) for different modulation schemes and coding rates and for both a macrocell and a femtocell base station. (CA = carrier aggregation; QAM = quadrature amplitude modulation; QPSK = quadrature phase shift keying) whereas the ­macrocell base station is installed outdoors; moreover, their antenna heights differ. We used the Erceg C and ITU-R P.1238 models, respectively, for the macrocell and femtocell base stations. An important parameter is the receiver signalto-noise ratio (SNR) for a certain bit error rate (BER) and depends on the modulation and coding rate used. The combination of the modulation scheme with the coding rate determines the physical bit rate, which is the total number of physically transferred bits per second, including useful data as well as the protocol overhead. The higher the bit rate, the higher the receiver SNR and the shorter the range.

Bit Rate We determine bit rate from several parameters, including the number of carriers containing user data, total number of carriers, bandwidth, modulation, coding rate, and, if applicable, carrier aggregation. This study considers only aggregation of component carriers with the same bandwidth. Note that the offered bit rate doesn’t differ between the macrocell and femtocell base stations because we consider the same technology for both base station types. 42

IC-17-05-Dery.indd 42

www.computer.org/internet/

Another important difference between the macrocell and femtocell base stations is the m ­ aximum number of served users. The smallest unit to which user traffic can be allocated is the physical resource block (PRB), which consists of 12 carriers. For each bandwidth, we divide the number of carriers used by 12 to determine the number of PRBs. F ­urthermore, we assume that each PRB has a different user. For the macrocell base station, this results in a maximum of 18, 36, 75, 150, 225, and 300 users, for a respective bandwidth of 1.4, 3, 5, 10, 15, and 20 MHz. For a femtocell base station, the number of users is typically limited to 16 (independent of the bandwidth),12 so we use it for the numerical results.

Results

We investigate the influence of carrier aggregation, heterogeneous deployments, and MIMO on energy efficiency.

Carrier Aggregation We investigate how adding carrier aggregation influences energy efficiency. Figure 2 shows the energy efficiency for some modulation schemes supported by LTE and LTE-Advanced for both macrocell and femtocell base stations in a 5-MHz channel. The energy efficiencies in Figure 2 noted as LTE apply to release 8/9, whereas all are applicable to release 10. The figure also indicates how much carriers are aggregated for LTE-Advanced. To obtain the results in Figure 2, we determine the power consumption, range, and bit rate as discussed earlier. The number of served users is fixed because we consider only one bandwidth (that is, 5 MHz). We then calculate EE using Equation 1. For each option in Figure 2 (for example, Macro LTE or Macro LTE-Advanced CA 2×5 MHz), a higher modulation scheme or coding rate results in lower EE because a higher modulation scheme and coding rate lead to a shorter range for a higher bit rate.2 For example, for an LTE-Advanced macrocell base station with a carrier aggregation of two component carriers, EE = 0.5 (km2 • Mbps)/W for a coding rate of 2/3 and a modulation of quadrature phase shift keying (QPSK) — that is, 2/3 QPSK versus 0.2 (km2 • Mbps)/W for 2/3 64 quadrature amplitude modulation (QAM). The reduction in range is greater than the increase in bit rate, leading to a lower EE as the power consumption and number of served users remains the same. IEEE INTERNET COMPUTING

12/08/13 3:26 PM

Designing Energy-Efficient Wireless Access Networks: LTE and LTE-Advanced

20 Macrocell base station Femtocell base station

18

Energy efficiency EE ((km2 . Mbps)/W)

16 14 12 10 8 6 4 2 0

1

10

Bit rate (Mbps)

100

Figure 3. Comparison of the energy efficiency of an LTE-Advanced macrocell and femtocell base station for different bit rates. Figure 2 shows that we can obtain higher bit rates even for higher EE using carrier aggregation. For example, for a macrocell base station and 1/2 QPSK, EE = 0.4 (km2 • Mbps)/W for LTE versus 2.1 (km2 • Mbps)/W when aggregating five  5-MHz component carriers. This is because carrier aggregation doesn’t influence the obtained range or number of served users. Introducing carrier aggregation has little impact on the base station’s power consumption because it corresponds in practice to a multicarrier (not aggregated) configuration that’s already supported by LTE or to a base station supporting multiple frequency bands, depending on the type of carrier aggregation.6 The extra power consumed for processing will thus be negligible compared to the base station’s power consumption. Thus, in terms of energy efficiency, it might be interesting to immediately implement LTEAdvanced in the network, without introducing release 8/9. We can obtain similar results for the other bandwidths.

Heterogeneous Deployments Figure 2 suggests that a femtocell base station is less energy efficient than a macrocell base station. However, this isn’t always the case. SEPTEMBER/OCTOBER 2013

IC-17-05-Dery.indd 43

Figure 3 compares the energy efficiency of LTE-Advanced macrocell and femtocell base stations as a function of attainable bit rates. To plot Figure 3, we first defined the bit rate for the aggregation of one to five component carriers with equal bandwidths. We considered all possible bandwidths. Next, for each possible bit rate and each base station type, we chose the most energy-efficient solution. Figure 3 shows that the bit rate determines which base station type is most energy efficient. For bit rates higher than 20 Mbps, the macrocell base station is the most energy efficient (EE = 7.5 (km2 • Mbps)/W versus 4.5 (km2 • Mbps)/W for 25 Mbps) due to its longer range and higher number of served users (despite its higher power consumption). Below 20 Mbps, there is no unambiguous answer. In some cases, the macrocell base station is most energy ­efficient (for example, for 5 Mbps), while in other cases the femtocell base station is most energy efficient (for example, for 13 Mbps). With carrier aggregation, we can obtain high bit rates even with a lower modulation scheme or bandwidth. For example, aggregating three 15-MHz carriers with 4/5 16-QAM gives us 122 Mbps, while aggregating five 20-MHz carriers with 43

10/08/13 2:52 PM

Feature: Wireless Networks

2.5

2.3

2.3

2.0

2.0

0.5 0.4

0.5

+393.1%

+327.2%

+203.5%

0.8

+331.3%

1.0

1×1 SISO

2×2 MIMO

4×4 MIMO

6×6 MIMO

8×8 MIMO

1.7

1.6

1.5

+121.3%

0.4

0.6

0.8

1.0 +131.3%

0.5 0.4

+94.3%

0.6

0.8

+103.5%

0.8

+217.7%

1.0

1.0

+304.8%

1.2

+270.9%

1.4

+47.2%

Energy efficiency ((km2 . Mbps)/W)

1.9

Macrocell base station Femtocell base station

1.8

0.2 0.0

in a more energy-efficient network. We can find this optimal combination by placing macrocell base stations to cover the area first, and then femtocell base stations to provide coverage in the coverage holes — that is, the areas that aren’t covered by macrocell base stations. Furthermore, we can extend the macrocell base station’s capacity as needed using femtocell base stations in the coverage cell of the macrocell base stations.

MIMO

2.0

(b)

+454.6%

1.4

1.4

+433.7%

1.8

1.5

0.0

2.5

+246.2%

(a)

Macrocell base station Femtocell base station

+94.8%

Energy efficiency ((km2 . Mbps)/W)

3.0

1×1 SISO

2×2 MIMO

4×4 MIMO

6×6 MIMO

8×8 MIMO

Figure 4. Influence of (a) spatial diversity and (b) spatial multiplexing on energy efficiency for different MIMO modes using 1/3 QPSK in a 5-MHz channel. (Percentages on the right side of each bar show the improvement compared to single-input, singleoutput [SISO]; the numbers at the top of each bar indicate the EE-values.) 2/3 QPSK gives us 113 Mbps, and aggregating four  20-MHz carriers with 1/2 16-QAM gives us 135 Mbps. Because of the higher modulation scheme, the range is much lower for 122 Mbps (141.3 m) than for 133 and 122 Mbps (respectively, 398.0 m and 327.2 m), resulting in a lower EE, which is responsible for the up-and-down behavior in Figure 3. Although future networks will obviously include different base station types, we will need a good estimation of the needed bit rate, coverage, and number of served users to determine which type is most suitable in each network location to reduce the network’s power consumption. Because demand varies over time, both base station types must be deployed in the network, with the optimal combination r­ esulting 44

IC-17-05-Dery.indd 44

www.computer.org/internet/

Figure 4a shows the influence of spatial diversity on energy efficiency for different MIMO modes in both macrocell and femtocell base stations. We consider a bit rate of 2.8 Mbps in a 5-MHz channel and assume that the mobile station has the same number of receiving antennas as the base station has transmitting antennas, which lets us determine the maximum energy efficiency gain.10 We use single-input, single-output (SISO) — that is, only one transmitting and one receiving antenna — as a reference scenario. Figure 4a shows that the more ­transmitting and receiving antennas, the higher the EE. For the macrocell base station, EE increases up to 433 percent when using 8×8 MIMO. In  ­Equation  1, Pel is 2 times higher, while R is 3 times higher, resulting in an EE that is 5 times higher. For the femtocell base station, EE increases up to 454.6 percent (or 5.5 times). Figure 4b shows the results for spatial multiplexing using 1/3 QPSK and a 5-MHz channel. Again, more transmitting and receiving antennas results in higher EE. For the macrocell base station, we get a maximum increase of 304.8 percent (or about 4 times) due to an 8 times higher bit rate, while the power consumption increases only 2 times. For a femtocell base station, the EE gain is a maximum of 131.3 percent (or 2.5 times). The highest EE gain is obtained using MIMO for spatial diversity. Future networks can benefit from MIMO, which currently has only limited deployment. Networks should use the highest possible MIMO mode whenever these bit rates or ranges are needed.

F

uture networks should implement LTEAdvanced, which will improve energy efficiency better than LTE. LTE-Advanced also adds enhanced support for relaying, which can IEEE INTERNET COMPUTING

10/08/13 2:52 PM

Designing Energy-Efficient Wireless Access Networks: LTE and LTE-Advanced

probably also improve energy efficiency. Furthermore, future work might include studying how the features investigated here will perform in terms of energy efficiency when applied on an actual network. Acknowledgments We conducted this work with the financial support of the iMinds project “Green Wireless Efficient City Access Networks” (GreenWeCan). Wout Joseph is a postdoctoral fellow of the Research Foundation Flanders (FWO-V).

References 1. M.W. Arshad, A. Vastberg, and T. Edler, “Energy Efficiency Improvement through Pico Base Stations for a Green Field Operator,” Proc. IEEE Wireless Comm. and Networking Conf., 2012, IEEE, pp. 2224–2229. 2. M. Deruyck, W. Joseph, and L. Martens, “Power Consumption Model for Macrocell and Microcell Base Stations,” Trans. Emerging Telecomm. Technologies, doi: 10.1002/ett.2565, 2012; http://onlinelibrary.wiley.com/ doi/10.1002/ett.2565/abstract. 3. C. Khirallah, J.S. Thompson, and D. Vukobratovi´c, “Energy Efficiency of Heterogeneous Networks in LTEAdvanced,” Proc. IEEE Wireless Comm. and Networking Conf., IEEE, 2012, pp. 53–58. 4. L.M. del Apio et al., “Energy Efficiency and Performance in Mobile Networks Deployments with Femtocells,” Proc. IEEE Int’l Symp. Personal Indoor and Mobile Radio Comm., IEEE, 2011, pp. 107–111. 5. 3rd Generation Partnership Project: Technical Specification Group Radio Access Network: Evolved Universal Terrestrial Radio Access (E-UTRA): User Equipment (UE) Radio Transmission and Recaption, TS 36.101 v8.17.0 Release 8, v9.1.0 Release 9, v10.1.0 Release 10, 3GPP, 2008–2011. 6. E. Dahlman, S. Parkvall, and J. Sköld, 4G LTE/LTEAdvanced for Mobile Broadband, Academic Press, 2011. 7. L.M. Correia et al., “Challenges and Enabling Technologies for Energy Aware Mobile Radio Networks,” IEEE Comm., vol. 48, no. 11, 2010, pp. 66–72. 8. J. Baliga et al., “Energy Consumption in Wired and Wireless Access Networks,” IEEE Comm., vol. 49, no. 6, 2011, pp. 70–77. 9. W. Vereecken et al., “Power Consumption in Telecommunication Networks: Overview and Reduction Strategies,” IEEE Comm., vol. 49, no. 6, 2011, pp. 62–69. 10. M. Deruyck et al., “Model for Power Consumption of Wireless Access Networks,” IET Science, Measurement, and Technology, vol. 5, no. 4, 2011, pp. 155–161. 11. I. Ashraf, F. Boccardi, and L. Ho, “Power Savings in Small Cell Deployments via Sleep Mode Technique,” Proc. IEEE Symp. Personal, Indoor, and Mobile Radio Comm., 2010, pp. 356–361. SEPTEMBER/OCTOBER 2013

IC-17-05-Dery.indd 45

12. M. Deruyck et al., “Modelling the Power Consumption in Femtocell Networks,” Proc. IEEE Wireless Comm. and Networking Conf., IEEE, 2012, pp. 30–35. Margot Deruyck is a research assistant in the Wireless and Cable (WiCa) research group in the Department of Information Technology  (INTEC) and Future Internet Department (iMinds) at Ghent University (UGent). Her scientific work focuses on green wireless access networks with minimal power consumption and minimal exposure of humans. Deruyck has an MSc in computer science engineering from Ghent University. Contact her at [email protected]. Wout Joseph is a professor in the domain of experimental characterization of wireless communication systems at Ghent University. His research interests include electromagnetic field exposure assessment, propagation for wireless communication systems, antennas, and calibration. Joseph has a PhD in electrotechnical engineering from Ghent University. Contact him at wout. [email protected]. Bart Lannoo is a postdoctoral researcher in the Internet Based Communication Networks and Services (IBCN) research group of the Department of Information Technology (INTEC) at Ghent University. As a member of the IBCN research group, he’s also affiliated with iMinds. His research deals with fixed and wireless access networks, focusing on MAC protocols, Green ICT, and techno-economics. Lannoo has a PhD in electrotechnical engineering from Ghent University. Contact him at [email protected]. Didier Colle is a researcher in the IBCN research group in the Department of Information Technology and Future Internet Department at  Ghent University  (INTEC UGent/iMinds). His work focuses on optical transport networks to support the next-generation Internet. Colle has a PhD in electrotechnical engineering from Ghent University. Contact him at didier.colle@intec. ugent.be. Luc Martens is a professor at Ghent University, and manages the wireless and cable research group at INTEC. His research interests include modeling and measurement of electromagnetic channels, electromagnetic exposure (for example, around telecommunication networks and systems such as cellular base station antennas), and energy consumption in wireless networks. Martens has a PhD in electrotechnical engineering from Ghent University. Contact him at luc.martens@ intec.ugent.be. 45

12/08/13 3:28 PM