Technology Vision 2020 Flatten Network Energy Consumption

Transcription

1 Nokia Networks FutureWorks Technology Vision 2020 Flatten Network Energy Consumption White Paper Nokia Networks white paper Technology Vision 2020 Flatten Network Energy Consumption

2 Contents 1. Executive Summary 3 2. Operators face rising network energy costs 4 3. A five-point approach to flatten network energy consumption 3.1 Base station efficiency Site optimization Network architecture evolution Network management and control Network modernization 16 4 Flat energy consumption despite accelerated traffic growth 5 Conclusion Page 2

3 1. Executive Summary Improving the energy efficiency of mobile networks will be a key challenge over the next decade as demand for mobile broadband services is expected to continue rising rapidly. Nokia believes that despite the expected huge growth in mobile traffic, it is possible to flatten the growth in network energy consumption and keep it broadly at today s levels, helping operators to maintain profitability. The focal point for improving network energy consumption must be the radio access network, because it accounts for 80% of all mobile network energy use. Fortunately, there are many opportunities to improve the energy efficiency in radio access, and Nokia has developed a comprehensive technological and architectural approach to address these. The approach covers improvements in five key areas: Base station efficiency more efficient baseband processing and highperformance radio front end amplifiers will boost energy efficiency. Site optimization lower energy costs by eliminating cooling and feeder losses, and implementing more renewable energy. Network architecture evolution improved resource utilization and capacity increase through multi-radio, densification, beamforming and distributed base stations. Capacity-driven network evolution will increase resource utilization avoiding idle nework elements. Network management and control teaching networks to be energy aware with advanced dormancy concepts.network modernization modernization at the right time to achieve optimum CAPEX and OPEX. The phase out of legacy technologies can be justified by energy savings alone. Nokia believes that achieving flat network energy consumption even in the face of huge increases in network traffic is a challenging, but achievable goal. This aim is embodied as one of the six pillars of the Nokia Networks Technology Vision Page 3

4 The Nokia Networks Technology Vision 2020 focuses on enabling mobile networks able to deliver Gigabytes of data per user per day, profitably and securely. Technology Vision 2020 comprises six technology pillars and paves the way for 5G: Supporting up to1000 times more capacity to meet accelerating data demand Reducing latency to milliseconds to prepare for the applications of the future Teaching networks to be self-aware and simplify network management by extreme automation Personalizing network experience to enable the business models of the future Reinventing telco for the cloud to create on-demand networks that are agile and scalable Flattening total energy consumption despite accelerated traffic growth 2. Operators face rising network energy costs Communication networks consume a significant amount of energy. Cellular networks in particular are estimated to be responsible for 0.5% of worldwide electrical energy consumption [1]. As most of the energy produced today is still generated from non-renewable energy sources, networks are correspondingly responsible for a significant amount of CO 2 emissions. Network energy consumption is also an important cost for operators. In mature markets, energy costs account for 10-15% of the total network operating expenses (OPEX) and can reach up to 50% in developing markets with a high number of off-grid sites, or where only a poor quality electricity grid is available [2]. The trend is upwards. In recent years the largest network operators reported a growth of 15-35% in their network energy consumption [3], the main reason being growing demand for coverage and capacity. Mobile networks have grown to support more than six billion subscriptions globally and traffic on those networks has been doubling every year, largely due to the increase in mobile broadband customers. 1. Fettweis and E. Zimmermann, ICT energy consumption trends and challenges, 11th International Symposium on Wireless Personal Multimedia Communications, Lapland, Finland, September ABI Research, Mobile Networks Go Green, Corporate Sustainability Reports, Multiple Network Operators, Page 4

5 Energy OPEX is 10-15% in mature markets Personnel expenses Site maintenance & rental Energy OPEX up to 50% in developing markets with high proportion of off-grid sites 28% 36% Mobile operator network OPEX distribution (example: Europe) Double digit growth in energy consumption in many networks Electricity 15% 9% 12% Others *Source: Capgemini, Operational Cost Strategies for Mobile Operators in Europe ABI Research, Mobile networks go green Backhaul Figure 1: Energy costs are a significant percentage of network operating expenses (OPEX) The global mobile broadband subscriber base is set to grow by 10% per year with mobile broadband penetration hitting 100% by 2020, up from 15% in Each mobile broadband subscriber will use an average of 25-50% more data per year. With this increase of traffic in the network, flattening the total network energy consumption, or even reducing it, will be a major objective for operators over the next decade in order to minimize adverse environmental impacts, meet emission targets, limit energy costs and stay profitable. 3. A five-point approach to flatten network energy consumption The radio access network is the dominant user of energy for mobile operators, accounting for some 80% of all mobile network energy consumption. Therefore, the radio access needs to be the focus of energy efficiency measures. Yet only 15% of that energy is used for forwarding bits, which means that 85% of the energy pumped into the network disappears and is not used for revenue generation. That s because most energy is used up by fans and cooling systems, heating and lighting, uninterruptible and other power supplies, and in running idle resources. Page 5

6 140 TWh -10% Energy transmission -10% 125 TWh NW Control Core Backhaul Base stations 80% -30% Cooling Heating Lighting UPS Storage loss -35% 90 TWh -20% Power supplies Fans -50% 70 TWh -70% Redundancy Idle resources -85% 20 TWh Energy from power plant Energy for telecom sites Energy for network elements Source: Nokia calculations based on published operator figures in 2012 Energy for chips UPS... Uninterruptable Power Supplies Figure 2: Where energy is consumed in a network - only 15% are used to transmit bits Energy for transmitting bits Clearly, there is real potential to reduce these secondary uses of energy in order to substantially improve a network s overall energy efficiency. Nokia has identified the key areas in which energy efficiency is ripe for improvement: Base station efficiency Site optimization Network architecture evolution Network management and control Network modernization 3.1 Base station efficiency In a simplified model the base station consists of the radio front end (RF) and the baseband processing unit. The RF part drives the antennas for transmission, and its efficiency is mainly determined by the efficiency of its power amplifier. The absolute power required by the base station is determined by the path loss of the radio signal, which strongly depends on the distance between the sending and receiving antennas, and on other propagation conditions such as reflections or obstacles. Page 6

7 For macro base stations the energy consumption is dominated by the RF frontend. Due to the small area to be covered, a single small cell requires significantly less RF power than a single macro cell. Therefore the energy consumption of small cells is more influenced by the baseband efficiency. The baseband module implements signal processing functions, and its efficiency is mainly determined by its digital processing efficiency. Therefore, baseband power consumption is not directly dependent on the cell s radius and relates more to the processing complexity of the respective radio technology. Base station average energy consumption Radio frontend efficiency RF mainly determined by power amplifier efficiency Baseband efficiency Macro cell BB Cell radius mainly determined by digital processing efficiency Small cell Relative dimensions just indicative RF BB RF: Radio frontend BB: Baseband unit Figure 3: Macro and small cells have different energy consumption profiles Page 7

8 60% 50% 40% 30% 20% 10% Peak PA line-up efficiency (3G/4G) LINC* Switched mode PA* Drain Voltage Modulation Envelope Tracking LDMOS 3-way/ Asymmetric Doherty HV-LDMOS GaN Doherty LTE Doherty PA Feed-Forward linearization 0% Digital Adaptive Pre-Distortion Smart Clipping Diamond* % 50% 40% 30% 20% 10% 0% Load independent efficiency Switched mode (2020) Doherty (2010) Resource muting Class A (2000) 10% Load level 100% New semiconductor technologies drive peak efficiency to the limits Peak power amplifier efficiency reaches a practical limit at ca. 60% New power amplifier architecture will further improve average efficiency Base station software mutes resources during low load conditions * Non mature technologies, research; LINC= Linear amplification with Nonlinear Components; PA = Power Amplifier Figure 4: Room for RF frontend improvement - from peak to average power amplifier efficiency Radio front end efficiency mainly determined by power amplifier efficiency The most important component influencing RF efficiency is the power amplifier. Figure 4 provides an overview of power amplifier efficiency evolution over the last decade as well as a future outlook. It shows that the state of the art system efficiency today is around 30% with gallium arsenide (GaAs) based Doherty power amplifiers. In the near term, asymmetric gallium nitride (GaN) based Doherty amplifiers will enable efficiencies beyond 40%, while also providing good wideband performance. Switched mode power amplifiers are expected to provide the next major evolution step, to take efficiency beyond 50%. The practical limit is expected to be around 60%, and this will probably also require new semiconductor technologies - such as transistors based on diamond substrate. These are in a very early research state today, and the coming years will show if and when they are suited for commercialization. So, unlike the rapid evolution of digital processing, the efficiency evolution of power amplifiers is a rather incremental and step-by-step process. Page 8

9 It is also important to note that amplifiers typically reach their highest efficiency only when operating at peak load. At lower loads their efficiency degrades rapidly due to a large idle power offset. Even higher efficiencies can be achieved by employing techniques such as Envelope Tracking, which adjusts the supply voltage of the power amplifier dynamically to ensure that it always operates at the point of optimum efficiency. On the downside, Envelope Tracking has only limited wideband capabilities. While this limitation is expected to diminish as a result of the previously described amplifier technology improvements, an energy-efficient network evolution will therefore always target to maximize the load of power amplifiers, for example by increasing utilization or by adding more spectrum to a site. In addition devices can be switched off in case of low load. Discontinuous Transmission (DTX) per symbol, called Micro DTX, is an example for such functionality. There is also promising progress on power amplifier miniaturization which will be a key ingredient of novel energy efficient base station concepts. Baseband efficiency mainly determined by digital processing efficiency Improvements in baseband processing will be especially important in boosting the efficiency of small cells. The smaller the cells get and the lower the required RF power, the more the power consumption of baseband processing affects the total power efficiency. With an increasing number of small cells relative to a macro base station, this effect additionally has to be multiplied by the number of small cells in the network. Fortunately, digital signal processing has seen rapid evolution. As formulated within Gene s law, the power dissipation of processors has halved every 18 months relative to their processing speed. In recent years this development has slowed because it becomes increasingly difficult to further reduce, for example, supply voltages. Nevertheless, the roadmaps of processor vendors show that over the coming years this factor will contribute to about 20-25% energy efficiency increase in processing annually, which is less than the gains in processing speed according to Moore s law. Smaller CMOS structures will also enable the integration of multiple functionalities into one System on Chip (SoC) and deliver efficiencies in hardware-based acceleration of baseband functionality. Of course the complexity of baseband processing will also increase over time as more advanced radio technologies are being introduced. Multi-antenna technologies such as 4x4 MIMO, coordinated multipoint transmission (CoMP) and advanced interference cancelation techniques will require significantly more baseband processing power. In turn it is expected that they will help to increase spectral efficiency by up to 10 times compared to the bps/hz in an HSPA network today. Page 9

10 Digital processing evolution Transistors [000] Clock [MHz] Processing power / Watt # Cores Baseband power efficiency does not follow Moore's law Restricted by constraints in gate capacities and input voltage reductions Smaller structures enable System-on-chip (SoC) integration and hardware acceleration Source: AMD, IBM, Intel processor roadmaps and datasheets Figure 5: Baseband: CMOS power efficiency gains stalled at 20% per year 3.2 Site optimization Multiple optimization opportunities exist at the cell site level. A key requisite is to develop and apply optimized site solutions, use renewable energy sources and advanced energy buffering, and optimize energy consumption to preserve the required site conditions. Site solutions: Every physical component of a base station site from length of cables, number of antennas, weather tolerance of the equipment, to the site layout itself impacts energy consumption. For example, using most efficient modern antenna solutions, sharing system modules for different radio technologies, deploying modern battery, rectifier and cabling products all contribute to energy optimization. Just think of the possibility of upgrading the performance or capabilities while saving energy consumption with Nokia s high-performance 6-sector site solution or with the synergies offered by the Single RAN and Single RAN Advanced solutions to possibly even reach Zero Emissions. The range of potential site optimization measures is very large and extends beyond the few examples given here. The key however is to tailor the site solution from the available options based on the specific conditions at each site. Page 10

11 The second important factor for site level optimization is to get the dimensioning right. The dimensioning needs to challenge traffic models and growth assumptions to improve the energy per traffic relation. It is quite usual that sites run over dimensioned for years, and might never reach high usage rates. Measurements and installation of smart meters will help to detect energy saving opportunities. Energy sources: Renewable energy sources such as solar power or wind power, and advanced battery technologies or fuel cells can further reduce the CO 2 impact of networks. The large scale aggregated battery capacity of a mobile network could even be used to save costs through the purchasing and buffering of energy when spot market prices are low. Energy consumption optimization: A site s energy supply solution should be designed for highest efficiency. Using high voltage DC or AC supply can result in 10% higher efficiency compared to 48 V DC solutions. What you don t know you can t improve: The prerequisite for site optimization is an appropriate measurement of the energy consumption. The ongoing standardization of test methods will help a lot to come to harmonized key indicators. If an optimization solution with energy as one of the performance indicators identifies a site resource, e.g. one of the small cells, that is not needed to meet the requirements of coverage, throuphput, loss or other parameters, then it can decide to shut down or mute these resources. Very effective is to avoid energy intensive cooling. Designs that can run at higher operating temperatures can eliminate, or at least reduce, the need for air-conditioning. If air-conditioning cannot be avoided, Nokia field studies (performed together with operators) have identified that there is significant savings potential by optimizing the cooling concept at a site. One example is the use of advanced battery technology: Cooling of lead batteries can consume a major part of a site s energy budget, whereas NiMH batteries can be well operated above 30 C and even then have a better life time. Page 11

12 Site solutions Energy sources Energy consumption optimization Local renewable energy as key for Zero Emission Solar panels or wind power with advanced battery buffering Fuel cells General Efficient energy supply solution Energy measurement Shutdown/muting of resources Modernize existing sites Leverage small cells Go for Zero Emission site Challenge traffic models and growth for dimensioning modernize Zero Emissions solution renewable energy CO2 6-Sector Antenna Indoor Airflow optimization Liquid cooling Outdoor (Flexi Multiradio BTS) No room, heating, air con Flexi Radio module for IP65 and up to +55 C without sun shield Top mast installation and Active Antenna Systems (AAS) reduce feeder losses Figure 6: Site optimization: eliminate cooling and feeder loss, add renewable energy This means that indoor sites shall adopt innovative approaches such as airflow and liquid cooling. The latter enables very effective cooling, supporting miniaturization as well as energy recovery with the heat transferred to an external heat exchanger and used for other purposes. The feasibility and benefits have been proven for the Flexi Radio module in a demo together with a major operator. Well-known measures such as consistently leveraging high-efficient power rectifiers all through the site will further drive down power consumption. Furthermore, in outdoor settings, advanced base station designs, as exemplified by the NSN Flexi Base Station, utilize highly efficient power supplies and support flexible shelter-less and fan-less installations, which radically reduce the overall energy consumption at a site. A further measure is to implement mast-top installations (RRH) that substantially reduce feeder length and therefore losses. 3.3 Network architecture evolution In addition to implementing energy-saving measures at individual sites, there is good potential for energy efficiency gains at the network architecture level. The need for more capacity will also significantly change the architecture of radio access networks over the next decade. Many of these changes will have an impact on the network s energy efficiency, or will even be driven by energy efficiency considerations. As a first step, legacy technologies should be phased out whereever possible in order to improve energy efficiency. Although a LTE base station consumes more energy than a GSM base station, it is far more efficient in terms of traffic per Watt. Page 12

13 When it is not possible to replace legacy systems, multi-radio deployments can play a key role. A single RAN base station with concurrent operation of GSM, HSPA and LTE enables the efficient sharing of resources between the different technologies. In the near term, when the load on the macro network increases, high-order sectorization will offer an energy-efficient upgrade path for the macro network. For example, upgrading a site to 6-sectors can provide up to 80% more capacity for the same total RF-power due to the higher gain of antennas with more focused beams, resulting in reduced interference. Active antenna systems (AAS) also support vertical sectorization and, in addition, avoid the massive feeder losses of conventional site designs. Smart antenna systems with adaptive beam forming can optimize this further, based on the current state of the network. Taken to an extreme, future technologies such as Full Dimensional MIMO (FD-MIMO) will deploy arrays with a multitude of small antennas for very fine granular beam steering. If it is possible to sharply focus the radio energy into small regions in space, this might lead to significant improvements in capacity as well as energy efficiency. When macro cell capacity upgrades reach their limits it will be an inevitable next step to introduce a layer of small cells to facilitate growth in capacity and high data rate coverage. As described earlier, a single small cell requires far less RF power than a macro site, given its limited cell radius. Due to the non-homogeneous spatial distribution of traffic in a network, small cells can provide an energy-efficient means to add high capacity in specific hot spots of a network. Another option for future radio networks is distributed base station architectures, in which the RF and baseband are separated into a remote radio head (RRH) close to the user and a baseband unit in a centralized Multi-radio: Single RAN base stations Phase out legacy technologies where possible Efficient sharing of resources by concurrent operation of GSM, HSPA and LTE Densification Adding capacity with small cells only where needed Increases average resource utilization Beamforming to increase capacity Efficiency improvements through interference reduction Active antennas and sectorization Smart beamforming and beam switching Distributed base stations Baseband pooling increases resource utilization Reduction of RF cabling losses Baseband RF1 RF2 RF3 Figure 7: Network architecture evolution: energy-efficient capacity with HetNets Page 13

14 location. Both are interconnected with a high capacity optical front-haul. This architecture increases energy efficiency by reducing coaxial feed line RF losses, given that the RRH is located close to the antenna. Centralized-RAN (C-RAN) takes this concept even further by pooling and sharing the baseband processing of multiple sites, allowing flexible resource assignment and higher utilization. Of course parts of these gains are offset by the additional energy consumption of the high capacity front-haul. Additionally there is potentially less hardware acceleration for C-RAN where the baseband pool is based on general purpose hardware. C-RAN is a suitable option for some operator use cases. Distributed base station architectures also simplify the implementation of distributed radio concepts such as cooperative multipoint transmission (CoMP), which can increase the network s overall performance, and thus its energy efficiency. With the experienced and expected traffic growth capacity-driven network evolution will allow more energy efficient use of resources, as increased resource utilization will improve the energy efficiency of the network. This avoides idle, unproductive resources that consume power. In many networks today the average utilization of base stations is very low (~5%), as most of the base station provide coverage and not capacity and therefore need to be operated independent of traffic. Our network evolution studies show that this average utilization can increase by up to 6 times as the traffic load on the network increases, the network granularity becomes finer and the adaptability to load conditions improves significantly. Network Utilization [%] % Busy hour Macro to the limits + HSPA micro Outdoor micro Indoor Pico/WiFi 30% Average utilization increase by factor 6 driven by capacity will enable efficiency improvements HSPA Today most cells for coverage, not capacity, resulting in a high portion of less used resources HSPA+ Average Densification Figure 8: Network architecture evolution: capacity-driven network evolution will allow more energy efficient use of resources Page 14

15 3.4 Network management and control With appropriate network management and control we can teach networks to be energy aware. Incorrect or sub-optimal configuration of base station parameters such as antenna tilt can quickly degrade the performance, and correspondingly the energy efficiency of the network. As networks become more complex, it becomes increasingly difficult to ensure optimal configuration of all network parameters. This challenge is addressed by the Self-Organizing-Network (SON) concept, which strives to automate large parts of the management of the network. SON will even enable dynamic adaptation of network configuration to optimize coverage and capacity in response to the current network state. Evolved SON will also enable advanced dormancy concepts to be realized. These facilitate direct control of network energy consumption through appropriate activation and deactivation of parts of the network in response to changing traffic load. Where average utilization of the network is low, the impact can be significant. The figure shows effective ways to intelligently adapt network energy consumption and avoid idle resources that consume energy. The chart shows the peak savings and the savings that were measured for selected features. According to Nokia measurements, operating resources in more efficient modes, muting or shutting down amplifiers or carriers during periods of low or no utilization offers a large savings potential. Advanced dormancy concepts Disable parts of the network based on time of day or load conditions Biggest impact for low load conditions and coverage part of the network Example levers to intelligently adapt network energy consumption network base station Shutdown of carriers and cells Switching off MIMO functionality for LTE Shutdown of a small cells Per symbol amplifier switch-off (micro DTX) Drain voltage modulation Peak savings W (GSM-LTE) 180 W 95 W Up to 40% (low load) Up to 25% (LTE) 150 GWh/a Network simulation* -11% -5% -1,6% 175 GWh/a Baseline *Source: Nokia analysis together with major European operator based on real network configuration Figure 9: Network management and control: Teach networks to be energy aware Page 15

16 3.5 Network modernization Finally it is important to address another factor - not technology-related - which heavily influences the level of energy efficiency that can be reached in a live network; namely the pace of network modernization. The gains in efficiency obtained through technology evolution materialize only when the new technologies are introduced into the live network. This effect becomes quite obvious when for example comparing the average power consumption of base stations in today s operational networks (e.g. as disclosed in sustainability reports) with the power consumption of a current state-of-the-art base station. Depending on the average age of equipment in a real network the state-of-the-art base station may consume less than half the energy of its operational equivalent. Determining the optimum timing for network modernization from a total cost of ownership (TCO) perspective requires of course careful balancing between CAPEX and OPEX. Figure 10 shows an example of the balance between modernization CAPEX and energy OPEX for a single base station. In a scenario with a very short modernization cycle energy costs are very low, as the latest equipment with highest efficiency is typically deployed in the network. However, the CAPEX resulting from modernization outweighs the savings in energy OPEX. In the opposite scenario, with an extended modernization cycle the additional energy OPEX outweighs the CAPEX savings from delayed modernization. In this example the calculation reveals an optimal point in a modernization cycle at around five to six years. CAPEX and energy OPEX vs. network modernization cycle TCO [k / 20 years] CAPEX TCO optimum OPEX Long cycle - Energy costs dominate Energy savings justify network modernization Assumptions: OPEX: 2.5kW initial energy consumption, 0.1 /kwh, 20% YoY efficiency improvements CAPEX: 10k BTS and site modernization costs, 5% YoY price erosion Modernization cycle [years] Figure 10: Network modernization: phase out legacy technologies Page 16

17 4 Flat energy consumption despite accelerated traffic growth Having outlined the areas in which energy efficiency can be improved, how much do they affect network energy consumption, despite growing traffic? Answering this question is especially important as many operators have publically committed to specific energy-saving targets for the future. Giving a general answer to this question that would be valid for all networks is not possible. Ultimately, the potential for optimization always depends on factors such as actual network configuration, equipment age, available spectrum, traffic patterns, geographic region, and more. For the sake of simplicity our analysis looks at an example radio access network that scales to a 100-fold traffic level. The model incorporates the key energy efficiency measures described within this paper to indicate the resulting improvements in overall network energy consumption. The model is based on a network with 20,000 macro sites and 30 million mobile broadband subscribers, and covers rural, suburban and urban areas. The traffic level in the network is scaled up from 200 MB (in 2010) to 20 GB (in 2020) average monthly usage per subscriber. The scenario also incorporates a capacity expansion scenario as would be expected in most mature markets. As traffic grows, the model network is expanded with new equipment to provide the required capacity. When expanding the network, the macro layer (36 dbm/mhz) is always evolved first, by upgrading from 3-sector to 6-sector sites, and by adding more spectrum. This is the most cost-effective evolution path, especially compared to adding new sites. When the macro upgrade path reaches its capacity limit (depending on available spectrum), outdoor Pico/ Micro cells (26 dbm/mhz) are additionally deployed. Other network scaling options such as WiFi offload are not modeled separately as they could be considered as deploying another type of small cell, and would not change the results significantly. To demonstrate the effects of network modernization, we assume the annual replacement of 20% of 2010 equipment with new state-of-the-art equipment. This corresponds to an equipment life-cycle of five years. To derive values representing energy consumption, we assume a mature network with an initial average power demand of 1.5 kw per site. For newlyadded equipment we take state-of-the art 2011 values for base stations equivalent to Nokia Flexi products, which are then extrapolated into the future to reflect technology progression. Figure 11 shows the resulting network energy consumption in relation to the traffic. Energy consumption of the considered network stays in the range of GWh/year. In this case the traffic is increasing by a factor of 100 and at the same time the energy efficiency increases by 110 times. This means that the energy efficiency can beat traffic growth, resulting in a reduction of the absolute energy consumption. Page 17

18 Base station efficiency: Reduce average power consumption Energy efficiency can beat traffic growth Network Power [GWh/year] x Energy Efficiency [TB/GWh] 200 Site optimization: Flexi Multiradio 10 BTS and Multicontroller Network architecture: Evolve to heterogeneous networks Network management and control: Teach networks to be energy aware Network modernization: Phase out legacy technologies x Legacy Macro Traffic Upgraded Macro Small cells 100x Source: Nokia analysis, example radio access deployment scenario with 20k Macro sites, 30m subscribers, 200MB/month/user initial traffic. 5 years equipment lifetime and rollout of key technology improvements & small cells Figure 11: Steps to reducing network energy consumption - Energy efficiency can beat traffic growth Over the first five years, efficiency gains are exceptionally large. There are two reasons for this. Firstly, the older, less-efficient equipment is replaced by much more efficient state-of-the-art equipment. Secondly, the average utilization of the network is significantly increased with growing traffic, especially during the first years. At 200 MB/sub/month, most of the cells provide coverage and are only lightly utilized, which leads to an average network utilization of around 5%. In 2015, with 5 GB/user/month, the average network utilization has increased by a factor of five to about 25% and reaches around 30% in 2020 with 20 GB/ user/month. To evaluate the impact of small cells on overall network energy efficiency, the calculation is repeated several times with different amounts of spectrum available to the macro layer. A small amount of macro layer spectrum results in a high number of small cells and vice versa. The results show that a denser layer of small cells can moderately increase the energy efficiency of the network. Of course this does not mean small cell deployments should be preferred over macro upgrades. Introducing a large number of small cells is associated with additional expenses related to site rental, backhaul and other costs, which cannot easily be compensated for by the potential energy savings. Therefore, from a total cost perspective, the most cost-effective approach remains to first expand existing macro sites as far as possible before deploying small cells on a wide scale. Page 18

19 5 Conclusion Solving the energy efficiency challenge for mobile broadband networks requires action at five layers. One part is a comprehensive technology approach that will significantly increase the efficiency of future radio base stations. As the base station is just one contributor to the overall energy bill, it is accordingly necessary to take a more holistic view, and seek to optimize energy efficiency at the site level and the network architecture level. In this context the future heterogeneous network architecture, involving multiple radio technologies and cell sizes provides both a challenge and an opportunity. On the one hand it creates increased network complexity. On the other, our model calculations show that small cells will offer an important lever to allow for continuous high-capacity and energy-efficient growth, beyond the limits of macro layer expansion. Energy aware software will enables radio access networks better adapt to variable network demand in different locations and at different times. On top of the technological and architectural efficiency improvements that we have outlined, there is an equally important need for a systematic approach to network modernization, to ensure that the promised efficiency gains actually materialize in the operational network. As the example of CAPEX/OPEX balancing shows, this step can actually be justified by savings from a total cost perspective alone. Last, but not least, the predicted data traffic growth itself will help to increase the energy efficiency of networks, by significantly increasing the utilization of network equipment. Taking all these factors into account, our studies indicate that keeping absolute network energy consumption essentially flat over the coming years, despite significant traffic growth is a challenging - but achievable goal. Page 19

20 Nokia is a registered trademark of Nokia Corporation. Other product and company names mentioned herein may be trademarks or trade names of their respective owners. Nokia Nokia Solutions and Networks Oy P.O. Box 1 FI Finland Visiting address: Karaportti 3, ESPOO, Finland Switchboard Product code C WP EN Nokia Solutions and Networks 2015