Final Report Summary - GREENNETS (Power consumption and CO2 footprint reduction in mobile networks by advanced automated network management approaches)
The report “Energy-saving solutions helping mobile operators meet commercial and sustainability goals worldwide” (Ericsson, 2008) and the article “Why the Telecom Sector Should Take Renewables Seriously” (Greentechmedia, 2013) both state that energy costs account for as much as half of a mobile operators’ operating expenses these days.
Radio network solutions that improve energy-efficiency are not only good for the environment; they also make commercial sense for operators and support sustainable, profitable business on mature market. The last statement is becoming even more important in the context of forecasts such as Ericsson’s Energy and Carbon Report (Ericsson, 2013), which predicts the total electricity consumption of mobile networks, including future wireless access points, to triple by 2020 in comparison to 2007.
The challenge has been recognised by the industry. Researchers, standard organisations and hardware vendors are successfully working on improved solutions such as better or smarter hardware, better software or communication protocols. All of this will contribute to saving energy in the future once introduced into the market at large scale. The GreenNets approach, in contract, is different.
The GreenNets project has looked into what can be done to save energy NOW. With the aim of cutting the energy consumption of GSM, UMTS and LTE radio access networks at least 10%, the project consortium focused on vendor agnostic approaches optimizing network operation. This strategy is based on two main observations:
• Current networks are typically designed and operated to secure coverage and capacity resources needed to deliver services in peak hours
• Energy consumption hasn’t been considered so far as radio access network optimization criterion.
The consortium has developed energy saving methodologies allowing for:
• Switching-off temporarily unnecessary (pieces of) equipment (e.g. GSM TRXes or UMTS cells), matching capacity to demand while keeping coverage in a multi-RAT environment and not influencing the network topology.
• Thinning out and optimizing topology of a single-RAT network by reconfiguration of antenna’s tilts, profiting from the same fact as described above.
The underlying idea of the proposed methods is to exploit coverage redundancies in order to adapt the network configuration to better match the actual service demands of the network. In such deployments there are usually in use technologies of different generations (2G, 3G, 4G) and different hierarchical levels (macro-, micro-, pico-, femto cells) with overlapping coverage. If the service demands can be fulfilled by different subsets of network elements the energy savings can be significant when configuring the network to having only one of those subsets of network elements active.
GreenNets solutions have been tested on the real network data. Implementing the first method in a single-RAT setting, just switching off redundant capacity, energy savings of up to 7,1% were uncovered. Moreover, such savings can be achieved within the current hardware and software landscape of mobile network operator. Designed as a functional extension of current planning tools and OSSs, the GreenNets software components can be easily fused with the planning, maintenance or optimisation processes as they use easily available performance and configuration data and are capable of automated implementation of changes in the network configuration (centralised SON) communicating with standard OSS interfaces.
Project Context and Objectives:
Energy efficiency is a top economic and political priority in Europe. This is reflected in the surge of activities in this area in the European Commission. On the 10th of November 2010, the European Commission adopted the Communication "Energy 2020 - A strategy for competitive, sustainable and secure energy", where it presented the new directions with respect to energy usage. The EU energy goals are also expressed in the "Europe 2020 Strategy for smart, sustainable and inclusive growth", as adopted by the European Council in June 2010. In particular, the EU aims at achieving ambitious energy and climate-change objectives for 2020: to reduce the greenhouse gas emissions by 20 %, to increase the share of renewable energy to 20 %, and to make a 20 % improvement in energy efficiency.
In case of mobile networks, the electrical energy costs constitute a significant share of operators’ operational expenditures (OPEX). They make up between 18% and 50% of operators’ OPEX. The radio access network part often contributes with a number as high as 60% of the total power consumption of mobile network operators. According to the analysis by The Yankee Group, the costs of energy spent on operating the radio network are approximately 2.6 % of the operator’s total expenditures (for the Polish operator PTC Era, 2.6 % of its expenditures in 2009 amounted to 26 Mio. EUR). Energy consumption is an increasingly important problem for communication networks. Already today it is not uncommon that operators become the biggest single power consumer in the nation, e.g. British Telecom consumed approximately 0.7% of the British energy during the winter of 2007, while the second biggest power consumer, Telecom Italia consumed 2 TWh in 2006.
The crude facts are that the present mobile networks are not energy efficient, that is, they consume more energy than required to satisfy the users’ demand. What exacerbates the problem is that they were not planned to be energy efficient. The roll-out focused on provisioning nation-wide coverage, capacity and consumers QoS demands.
The increase of energy efficiency in radio networks is expected to come from the following lines of developments, 1) Energy efficient equipment (out of scope of this project / domain of equipment vendors), 2) Energy efficient network operations, and 3) Energy efficient radio deployment.
For the second and third line, the idea is basically to operate just as much of the radio network as needed. In both cases, the reduction in energy consumption will largely arise from a smarter use of existing and new network equipment. This is strongly advocated by the Next Generation Mobile Networks Alliance (NGMN) in the document on Operational Efficiency (P-OPE). Out of the 10 main approaches, increasing energy efficiency is listed as 3rd most important. As the energy is a main part of the operational expenses, not only network elements with low power consumption become more and more important but also the temporary shutdown of unused capacity is valuable and explicitly mentioned in the document. With energy efficient network operations, this goal is pursued by adapting the use of the radio infrastructure to the current demand. For example, the unnecessary parts can either be put into sleep mode or switched off. However, an automatic capacity-driven energy saving mode can only be realized in existing networks using network management systems based on performance data. The possibility to temporarily switch-off (parts of) radio access network nodes, e.g. for a given Radio Access Technology (e.g. GSM, UMTS) is expected to reduce the operational costs related to power consumption. The case of energy efficient radio deployment goes one step further. The objective is to explicitly develop a network infrastructure (i.e. migrate an existing infrastructure over a time span of years) that is particularly well prepared for energy efficient network operations in the sense described above. Hence, the gain from an energy efficient network operation is expected to be considerably larger when the network layout is prepared for this.
The following 6 main project objectives are targeted:
S&T Objective 1: Utilize tools from stochastic modeling to develop reliable models for energy consumption and the CO2 footprint of cellular networks as well as user mobility, load distribution, and interference.
S&T Objective 2: Exploit hierarchical architectures of wireless cellular networks for energy savings by reconfiguring a wireless access network based on daytime-dependent load fluctuations.
S&T Objective 3: Develop algorithms for dynamic load-dependent selection and reconfiguration of radio access technologies.
S&T Objective 4: Demonstrate the feasibility and benefits of the pursued approach by implementing and integrating the developed algorithms on the target network management platform as optimization engines and by validating them on an actual operating network.
S&T Objective 5: Develop flexible, efficient and powerful simulation platform for modeling radio networks of different types and technologies to reliably estimate their energy consumption in the context of continuously rising data rate requirements.
S&T Objective 6: Support the evolution of radio networks towards energy-efficient networks based on the optimization of the network deployment.
Stochastic models for energy consumption
A high level power consumption model was developed which, in addition to the radio equipment, also includes auxiliary site hardware such as climate control or microwave backhaul. We provide a detailed description of the model as well as exemplary results for different hardware generations and radio access technologies (RAT). The developed model was implemented to be used in the Energy Efficiency Monitor (EEM) as well as part of the Network Simulator. Smart meter measurement data provided by BENCO was used for plausibility checks and allowed for a qualitative evaluation of the modelled results.
Figure 1 Spatially mapped voice traffic [Erl] for one hour period
Stochastic modelling of user and load distribution
Furthermore, algorithms to map network key performance indicators (KPIs) in a spatial and temporal varying manner to a macroscopic scenario were developed and refined during the project, see Figure 1. The developed mechanism was implemented to form a part of the Energy Efficiency Optimizer (EEO). The pixelmaps created by this component form one of the required inputs for the EEO.
As we have identified, the one most promising approach to save energy is to switch cells off when traffic demand is low. Before devising algorithm for energy savings, we first analysed the impact of such configuration changes on the network performance. Unfortunately, deriving parametric models to predict most KPIs as a function of the network configuration has been proven difficult, and the sheer number of KPIs collected at cells further exacerbates this problem. As a result, instead of devising models for each individual KPI collected by operators, we determined the influence of switching cells on or off by studying measurements of real running networks where sites had been introduced. To this end, we used state-of-the-art statistical tools to solve hypothesis testing problems that have the objective of indentifying, with strong statistical guarantees, cells and KPIs that are likely to be impaired by introducing and powering down whole sites in the network. We note that for this analysis we used real measurements from GSM and UMTS networks.
Network data collection for energy efficiency
We developed algorithms for two main subtasks: data collection with reduced communication overhead and voice traffic forecast in 2G and 3G networks.
In more detail, in the first subtask we have developed feedback techniques based on principal component analysis (PCA) to reduce the dimensionality of a multivariate data set, which, here, is the set of the KPIs reported by base stations. These techniques can greatly decrease the overhead in the network, which allows base stations to report KPI measurements more often than currently possible. As a result, having more information about the network state, algorithms for energy optimization are able to make reliable decisions. We note that configuration management (CM) and performance management (PM) data observed in the network is huge, but in many cases the network behaviour can be described by a set of transformed variables having lower dimensionality compared to the original set. PCA is able to determine such transformed variables in a mathematically rigorous way. In order to have (at least approximately) independent and identically distributed samples, data filters have been implemented to pre-process the data before applying PCA. Such filters include daily averages (excluding weekends and holidays) and selections of measurements taken at given hours and/or days (e.g. all measurements taken on Mondays between 9am and 12). Our empirical evaluation, which used a large dataset consisting of KPI measurements obtained in a real network, showed that a reasonably accurate estimate of the original KPI values can be obtained by using less than one half of the original dimensions (i.e. the original number of KPIs). In addition to decreasing the overhead in reporting the network status to operators, the output of PCA can also be used to detect changes and anomalies in the traffic, globally or in a given site. Statistical tests based on subspace distance, or the squared prediction error (SPE), can be employed for this application.
In the second subtask, we developed machine learning tools to produce forecasts of KPIs that have a coarse periodic pattern. Typically, such KPIs have a strong correlation with voice traffic, which presents a periodic pattern where reported values are large during the day and low at late night. These prediction tools are one of the key building blocks of the energy saving optimization techniques developed in the project. More precisely, with current technology, changes requested by operators are not implemented immediately; large delays are expected, and energy saving tools should act based on forecasts in order to give operators enough time to implement the required energy-saving configuration changes. One of the main features of the proposed techniques is that they are able to include contextual information such as the presence of holidays and weekends, in addition to providing network operators with confidence intervals for the predictions. Figure 2 illustrates a typical output of the prediction tool.
Figure 2 Predicted and observed values of a voice-related KPI in a real GSM network.
Information theoretic bounds for energy consumption
We investigated how the energy consumption and throughput scales with the traffic demand in a wireless network. In particular, we focused on the asymptotic regime of a large-scale random wireless networks where the number of users, which are placed arbitrarily at random over some area, tends to infinity. We showed that the throughput per user tends to zero unless the number of base stations scales linearly with the traffic demand. Unfortunately, such a base station deployment strategy is infeasible with respect to installation costs and energy consumption. However, the analysis provides valuable insight into the rate at which the cellular infrastructure needs to grow to provide a good compromise between per user throughput and energy efficiency. In this respect, we identified an optimal infrastructure node scaling rate for which the throughput only moderately tends to zero, and we obtained a good energy-per-bit scaling behavior. Figure 3 illustrates our results in this task.
Figure 3 Throughput per node as a function of the number of users.
Simulation platform for evaluation the energy consumption
Network simulation capabilities were developed to be able to simulate realistic network deployments with the temporally and spatially varying traffic distributions that were developed in the project. To allow the reuse of already available simulation capabilities offered by typical system-level simulators such as atesio’s NGPS platform, the developed simulation platform is formed by distinct extensions that upgrade the capabilities of such a simulation core, as depicted in Figure 4. One example is the calculation of a network’s power consumption through a dedicated extension that reuses code originally developed for use in the EEM and bases upon load calculations from the system-level simulation core. Throughout the project, additional simulation results were identified to be of special interest to evaluate multi-technology networks and the developed optimization algorithms. Resulting from literature studies, suitable approaches to simulate the blocking probability and the impact of different antenna configurations, e.g. MIMO and scheduling schemes were chosen and implemented as distinct extensions. An additional extension was developed to allow the distribution of user traffic among different network layers and technologies while regarding service-specific load constraints.
Each extension builds upon the capabilities of the underlying system-level simulation core and communicates with it via a dedicated interface.
Figure 4 Basic structure of GreenNets Network Simulator
Map based prediction using machine-learning approaches
We have investigated to which extent machine learning tools are able to identify signaling conditions that can be associated with a high probability of handovers. To this end, we constructed, by means of simulations, a dataset with handover positions and their corresponding signaling conditions. This dataset was then used as an input to machine learning tools based on logistic regression and Gaussian processes (GPs). The output of such classifiers is a map showing how likely a handover should occur at a given position in the map. Figure 5 depicts a typical outcome of the algorithm. In particular, we showed that GPs can perform slightly better than logistic regression for this task, but at the cost of much larger computational complexity.
Figure 5 Exemplary probability map obtained with the logistic regression algorithm. Values close to one indicate areas with high probability of handovers
Self-optimization of capacity – coverage – energy consumption tradeoff
Starting from a general model that can be applied to technologies with fast link adaptation (e.g. LTE), we developed algorithms that are able to decrease the energy consumption of cellular networks while meeting a target minimum quality-of-service (QoS). In this study, QoS stands for the minimum throughput required by users, which is estimated by using the developed traffic forecast tools. The main idea is to identify network elements (sites or individual sectors) that can be switched off. We showed that, typically, the resulting optimization problems are NP-hard, so we cannot expect to solve it both fast and optimally. Given the sheer size of the optimization problems, we opted for fast heuristics that builds upon recent techniques that appeared in the compressive sensing literature. More precisely, after suitable convex/concave relaxations, we arrived at continuous optimization problems that, although still intractable if optimal solutions are sought, can be addressed with majorization-minimization (MM) techniques. In addition to the computational speed to obtain an approximate solution, the proposed techniques also have a strong analytical justification, which is a desirable feature that is absent from many algorithms proposed in recent years. We also showed that our general scheme can be applied to current standards, and, in particular, to GSM, UMTS, and LTE networks. Figure 6 illustrates a typical output of the algorithm, which is the core of the energy efficiency optimizer (we note that in this figure we show results based on data obtained from a real network deployed in a major European city.)
Figure 6 UMTS sites that can be switched off based on the predicted traffic at 1AM. Colors indicate traffic intensity (blue: low traffic red: high traffic). Red circles represent sites that should remain active, while blue circles represent sites that can be switched off. Most of the map has been pixelated owing to nondisclosure agreements between the consortium partners and the operator that contributed with the dataset.
Energy-efficienct Network Topology Configuration
This task had the objective of devising techniques for fast assessment of configurations and for network planning (e.g. antenna tilt optimization). The results of this task are the core of the “network designer”. In more detail, we show algorithms that have the objective of saving energy in current networks (i.e. networks using existing technologies) by exploiting the well-predicted behavior of traffic in a typical day or week. Those algorithms are based on optimization techniques that, for mathematical tractability, assume the worst-case interference scenario. By doing so, the algorithms are remarkably conservative; the solution to the above optimization problem assumes that the quality-of-service (QoS) requested by users can be supported even if every base station experiences the maximum expected interference from neighboring cells. In practice, however, the load is smaller than that used in the above optimization problem. As a result, there is a great potential to save further energy and to improve the network (in terms of, for example, total system throughput) if the actual load values are used by the optimization algorithms. In light of this background, we cite as the main contributions of this task:
i) Development of load computation techniques for systems with fast link adaptation. We showed that recent schemes for load computation in LTE systems can be extended and unified by using the mathematical framework of interference calculus. In particular, we devised iterative schemes that can compute load under a plethora of assumptions that are often overlooked in the literature. As a particular example, our schemes can take into account the maximum spectral efficiency of practical modulation techniques. Despite the fact that the convergence of the iterative schemes is only asymptotic, the schemes are able to provide information about the accuracy of the load estimates at each iteration. They can also be used to compute the fraction of the requested traffic that can be served by a given network configuration (if the demand is too high), which is a problem known as “load control” in the literature. Mathematically, the load control problem is special case of a complementarity system, and we proved that this special system can be solved with the proposed iterative schemes.
ii) We combined the above results on load estimation with the optimization algorithms. By doing so, we avoided the worst-case interference assumption. In an extreme numerical experiment assuming lightly loaded networks in a dense urban environment (which could represent a real network at late night), we reduced the energy consumption to about 25% of the energy consumption of the configuration obtained by using only the energy saving algorithm.
iii) We further extended the above energy-saving scheme with realistic load computation by assuming that not only sites can be switched off, but by also assuming that antenna tilts (electrical and mechanical) can be adjusted. This additional degree of freedom further increases the energy saving potential, but it also greatly complicates the optimization problem, which is already NP-hard in its simplest form. However, we demonstrated that these techniques and in item ii) above can be combined to cover the scenario where tilts can be adjusted. We evaluated these extensions by using the simulation platform NGPS developed by one of the consortium partners (atesio).
iv) As a secondary objective of this task, we provided the SMEs with general ideas to extend even further the proposed mathematical tools to consider future systems where, for example, relays are present.
Decision Algorithms for Selection and Configuration
In the first part of the project we primarily focused on energy saving schemes for individual technologies. In contrast, this task introduced novel algorithms for multi-RAT systems that have the objective of reducing the total energy consumption of heterogenous networks. The tacit assumption used in this work is that users can be served by multiple technologies, and the technology choice can be left to energy saving algorithms (which should guarantee the users’ minimum data rate requirement and coverage). In more detail, we combined the individual constraints for GSM, UMTS, and LTE systems into a single optimization problem similar to that developed in the same deliverable. By doing so, the optimization tools which considers the worst-case interference scenario and the ones which considers scenarios with accurate load levels became available to solve the resulting optimization problem. We performed extensive numerical experiments in synthetic networks to validate the good performance of the multi-RAT optimization algorithms.
Network reconfiguration triggering
The energy saving algorithms described above have mostly considered traffic at particular time instants. More precisely, for a given fixed traffic the algorithms are able to provide operators with a network configuration that is able to provide the necessary quality-of-service to users with little energy consumption. However, in real networks, traffic is constantly evolving. As a result, the energy optimization algorithms are faced with two additional tasks that have not been covered before; namely, i) when to reconfigure the network, and ii) what traffic volume should be considered in the optimization process in order to guarantee the necessary quality-of-service for a sufficiently long time. In light of this background, we first described how the above-mentioned tasks can be performed in real networks. During the development of this technique, we also addressed one of the main limitations of the prediction schemes outlined earlier, which is the fact that those forecast tools have good prediction power only for KPIs with a coarse periodic behavior. Unfortunately, this property is not apparent in many KPIs that are required as an input to optimization algorithms, such as KPIs related to data traffic. As a result, the second objective of this task was to devise prediction tools for irregular KPIs. To this end, we proposed robust approaches based on results in order statistics, and we evaluated the performance of the prediction tools by using a large dataset consisting of historical traffic measurements of a real network in Europe.
Definition of optimization engine architectures and software interfaces
We defined the architecture of the EEO, including its software building blocks and interfaces to existing platforms of SMEs: RANOOK and NGPS. The EEO was designed as an autonomic system, capable of proposing and executing actions (e.g. proposing the times to switch network elements off and on) that can lead to energy savings in the network.
Furthermore, the EEO was designed to be vendor-agnostic solution that may cooperate with any cellular network platform. To this end, the EEO communicates directly with a component called Data Abstraction Layer (DAL), also designed and developed within this work package. The goal of DAL is to hide heterogeneities of the data sources and platforms, and to provide a unified and managed read-write access to these sources. In particular, within this task, we defined appropriate APIs that allowed DAL to communicate with RANOOK and NGPS.
A functional overview of the EEO is given in figure below.
Figure 7 General overview of the Energy Efficiency Optimizer communicating with data sources via Data Abstraction Layer.
Implementation of algorithms
We implemented the EEO product as consisting of the Manager and three Engines: the Prediction Engine, the Pixel Map Engine, and the Optimisation Engine. The Manager is the heart of EEO and runs as the main process of the EEO. It overlooks the data flow between Engines and communicates with various sources via DAL.
However, the main focus of this task was to provide efficient implementation of the Engines. The prototypes for these engines have been developed in preliminary tasks. The three Engines are mathematical, reusable libraries that perform certain type of computations and optimizations. The engines were programmed with special emphasis on efficiency, numerical stability and accuracy, and—whenever applicable—reproducing the exact outcomes of the Matlab prototypes designed in previous work packages.
The overview of the software building blocks of EEO is given in figure below.
Figure 8 Three Engines of the EEO
Integration and benchmarking
We thoroughly tested the developed Engines and the whole EEO product using test cases. The test cases for Engines covered correctness tests, timing tests assessing efficiency of the programming solutions, and compatibility tests ensuring that the developed solutions return same or similar results to the prototypical implementations delivered by HHI and TUBS.
The large part of this task was however integrating the Engine in the complete, working EEO product, and integrating the EEO with DAL and RANOOK and NGPS. The former required ensuring a proper data flow inside the EEO and the latter involved programming of special APIs allowing DAL to access RANOOK and NGPS platforms.
The socio-economic impact
During the past three years there are clear trends that energy consumption reduction became a strategic goal for mobile network operators (MNO) around the globe. Different sources name a bit different numbers of potential financial losses, waste of energy or CO2 footprint, but all of the sources proof the critical need of improvements in the area of mobile network energy consumption optimisation.
According to press information report “Energy-saving solutions helping mobile operators meet commercial and sustainability goals worldwide” (Ericsson, 2008) and the article “Why the Telecom Sector Should Take Renewables Seriously” (Greentechmedia, 2013) energy costs account for as much as half of a mobile operators’ operating expenses. Additionally Ericsson’s Energy and Carbon Report (Ericsson, 2013) forecasts the total electricity consumption of the ICT sector to increase by almost 60 percent from 2007 to 2020 owing to the increasing number of devices and network expansion. As for the electricity consumption of mobile networks, including future wireless access points, is expected to not more than triple by 2020.
Thus radio network solutions that improve energy-efficiency are not only good for the environment, they also make commercial sense for operators and support sustainable, profitable business. There is a clear trend that functionality of mobile phones increases dramatically over the past two decades, transforming the mobile phone to a device capable of far more than simple voice calls. Market demands are expected to increase the bandwidth of the digital cellular network up to 8 times over the next years. That is possible by installing additional equipment (e.g. 3G and LTE layers as capacity over existing 2G or 3G as coverage layes) or adding more sites with reduced heights and cell sizes, which will lead to even higher total energy consumption. The report highlights the MNO focus on OPEX optimisation as well. Improvements must be achieved by complex solutions combining new generation of radio equipment amended with additional control solutions.
Another report published by Capgemini (2009) analyses various cost reduction strategies for MNOs and network operating expenditure is identified as one of the key areas for improvement, since operating costs take from 50% up to 75% of MNO revenue. The network operation expenses makes the 23.7% of the overall OPEX and electricity costs makes 15% of the network OPEX. The report also states that energy costs are driven largely by sites’ cooling, thus solutions for cooling efficiencies improving and reducing energy consumption would realise a tangible savings. These measures also ensure that operators make a meaningful contribution towards the environment, reduce the CO2 emissions and benefit from the corresponding goodwill generated. Capgemini report also forecasts cascaded energy savings due to possible application of GreenNets or other similar technology GreenNets solution will target radio standby mode contributing to app 62.5% of energy consumption share and later the usage of advanced climate control. Passive or free cooling is being introduced by operators at the moment, thus this is not part of GreenNets solution.
The findings of the Capgemini’s report are confirmed by the latest resources as well, for example the article “Saving Operating Expenses in the Mobile Backhaul” (Juniper, 2013) states the network operating costs are even higher and reach 30%, since the expansion of the 4G/LTE services requires a substantial increase in the network bandwidth in terms of extra equipment and power consumption.
Leading network equipment providers (NEP), supplying MNOs around the globe, are proactively tackling the same issues too by running their own economy studies, announcing the corporate commitment for the energy and CO2 footprint reduction and developing modern energy effective offerings. Numerous initiatives and solutions are popping up at the mobile operators market during last 12-24 months and there is no unified trend how to address the problem at the moment.
According to other sources, the mobile network 80% to 86% of a mobile operator’s energy and from two thirds to three fourths of that is for the equipment on sites (Global Telecom Business magazine, 2010, Nokia Siemens Networks, 2012). In mature markets, up to 10% of network OPEX is used on energy, but in developing markets, it can be from approximately 15% up to even 30% of networks OPEX (Nokia Siemens Networks, 2012). The expenses for electricity ranged between 18 % and 50 % of operators OPEX expenditures - the numbers are highly dependent on the market type and to the lesser extent on the novelty of the operator’s infrastructure. Energy-saving solutions helping mobile operators meet commercial and sustainability goals worldwide. It results that electricity bill of mobile network equipment over its lifetime is relatively very high and the accumulated cost can exceed the cost of equipment (interview with R. Ryliškis, OMNITEL Lithuania).
Additionally to high expenses, electricity overuse results in unreliable power supply in certain locations. According to report “Establishing efficient power & environmental monitoring systems” (Huawei, 2012) the 65% of communications interruptions are caused by power supply failures and 85% of them discovered after more than 12 hours thanks largely to customer complaints. Both MNO and NEPs are announcing their commitments to reduce the energy consumption and CO2 footprint as part of the companies’ strategic targets. For example, Vodafone puts a target of 50% CO2 reduction target in developed markets versus the 2007 baseline by March 2020 in its annual report (Vodafone, 2012). Vodafone Turkey and Huawei have published the case study about the significant optimisation of energy consumption with 355k USD savings per year due to updates on up to 1500 base stations in Turkey (Vodafone Turkey, 2010). TeliaSonera also highlights that one of the permanent priorities is to continuously try to find more energy-efficient solutions for networks and data centers in its annual report (TeliaSonera, 2012) and corporate responsibility and environment statements (TeliaSonera, 2012). All major European MNOs as members of industry organisations like European Telecommunications Network Operators' Association (ETNO) and European affiliate of Groupe Speciale Mobile Association (GSMe) are following the EU Commission recommendation (2009) encouraged intensified efforts from the sector in the transition to a low-carbon economy.
Mr Rolandas Ryliškis (Vice President of Omnitel, TeliaSonera, Lithuania) confirms these trends and market demand in his interview, conducted in terms of this document. Energy consumption reduction is one of the main strategic goals for the Omnitel and other MNOs in Baltics. Specific annual targets and responsibilities are introduced both on company and Operation Departments’ levels. According to R. Ryliškis network operations costs optimisation a key reason for the energy consumption reduction. Although there are serious additional social responsibility reasons as well:
• Reduction of CO2 emission is a hot topic both in Lithuania and in Europe. There is even dedicated ISO process for CO2 emission monitoring and reduction introduced in TeliaSonera,
• Reduction of electromagnetic radiation level in residential areas or workplaces is also actual problem in some countries. For example this issue is heavily exaggerated in Lithuania as a result of political communication, but nevertheless it is real and must be faced by technology.
Last, but not least problem related to energy consumption management is the need ensure reliable base network even in the conditions of instable power supply or emergency weather conditions (storms). In this case the focus is not on the reduction of costs, but on prolonging the lifetime of the core network service provided by the site, while temporarily running on batteries or insufficient energy supply with auxiliary services shut down remotely.
Here we have reviewed only part of the publicly available and official resources proving the highest demand for such products as GreenNets world-wide. The global scale and massive involvement of different parties in the enlisted initiatives also proves our strategy to go global with selected innovative B2B product with the help of cross-country partner network.
Main dissemination activities
A dedicated GreenNets web portal has been set up in November 2011 and is available under the following link: http://www.greennets.eu/. The portal serves as a dissemination channel for the GreenNets project, including publication of any project relevant information. Furthermore, it provides for a working environment for GreenNets partners with dedicated collaborative tools in form of discussion forums with easily manageable threads and possibility to upload documentation. Threads have been set up to archive project documentation and support collaboration within work packages. Furthermore, dedicated threads are set up for key cross project activities and to store material from the project meetings. Moreover, a dedicated space has been set up to facilitate communication with the External Advisory Board in a form of an interactive forum.
A project movie showcasing the issue of power consumption in mobile networks and the associated challenges to be solved by the GreenNets project was prepared and released on the 7th February 2012 on the GreenNets portal and on YouTube.The movie is available under the following links:
- VTC Conference, Dresden, Germany, 2-5 June 2013;
- The 7th IC1004 MC and Scientific Meeting and the Joint COST IC1004 + GREENETS Workshop on “SON Algorithms for Energy Efficiency; Ilmenau, Germany; May 28-31, 2013.
- Szymon Stefański: “ Mechanizm prognozowania natężenia ruchu w sieciach radiokomunikacyjnych dla rozwiązań poprawiających efektywność zużycia energii” Przegląd Telekomunikacyjny i Wiadomości Telekomunikacyjne nr 6/2013, prezentowany KKRRiT, Wrocław 06-2013,
- Johannes Baumgarten, Andreas Eisenblätter, Thomas Jansen, Thomas Kürner, Dennis M. Rose, Ulrich Türke, SON Laboratory: A Multi‐Technology Radio Network Simulation Environment for the SON Analysis (Demonstration),IWSON 2012 (International Workshop on Self‐Organizing Networks), August 2012, Paris,
- GreenNets in promotion of the 7th Framework Programme conducted by Wrocław Centre of Technology Transfer
- Press release - http://www.telepolis.pl/news.php?id=23992
- GreenTouch Meeting, Sttutgart, Germany, Nov. 8
- R. L. G. Cavalcante, E. Pollakis, S. Stanczak, S. Stefanski, R. Nowak, T. Kürner, A. Eisenblätter, and D. Montvila, "Energy Savings in Cellular Networks," in Proc. COST IC1004 + GREENETS Workshop on “SON Algorithms for Energy Efficiency, 2013.
- R. L. G. Cavalcante, S. Stanczak, A. Eisenblätter, T. Kürner, D. Montvila, R. Nowak, and S. Stefanski. "Voice and Traffic Prediction in GSM and UMTS Cells with Real Network Measurements." (2013).
- E. Pollakis, R. L. G. Cavalcante, S. Stanczak, “Reducing the energy consumption of multi-RAT cellular networks with a sparsity promoting algorithm,” IEEE Transactions on Wireless Communications, in preparation
- E. Pollakis and S. Stanczak, “Energy-per-bit scaling laws of large wireless networks with infrastructure nodes”, IEEE Transactions on Wireless Communications, in preparation
Exploitation of results
Atesio intends to customize several distinct products (software as well as services) on the basis of their next generation planning system (NGPS). NGPS is a radio network analysis, planning, and optimization platform that is used by atesio in a wide range of commercial projects. The platform provides highly efficient, highly integrated and fully automated planning and optimization functionalities for (mixed) GSM/EDGE, UMTS/HSPA, and LTE radio networks. On the one extreme, the platform is used within a traditional planning framework as well as for rapid large-scale network analysis and optimization as needed for strategic network development. A seamlessly embedded concept of configuration states allows for maintaining alternative (or consecutive) network states as to model planning alternatives and network evolution. On the other extreme, its detailed network models paired with its ability to use measurement data of various sources (e.g. PM data) make it available for a SON-like applications when integrated with a network configuration management system. This wide range of applications and its flexibly combinable planning and optimization functionalities render NGPS a strong building block for highly integrated and automated network operations environments. NGPS functionalities can be called as software services; it can be run on centralized, distributed servers, or in a hosting environment.
As the result of the project, the NGPS platform shall be extended such that the following use cases become available on the platform:
• Business Case Demonstrator
• Network Simulator
• Demand Forecaster
• Network Designer
The degree to which the other two use cases, namely, Energy Efficiency Monitor and Energy Efficiency Optimizer will be supported is not yet determined. At the time of writing, the NGPS platform is not interfacing live (online) with CM and PM. Such an interfacing, however, is a prerequisite to strongly support these use cases. Functionality that is not explicitly dependent on this interfacing shall be made available. With respect to Energy Efficiency Monitoring, for example, the analysis may not be live in the sense of a control panel. But the analysis functionality based on historic and recent PM data is nevertheless valuable and shall be made available. The reason for this is simple, even in retrospect the need for improved energy efficiency may be demonstrated very well
BENCO aims to support and consult the development of RANOOK by DATAX and NGPS by atesio so their product solutions can interface with existing and upcoming Smart Metering solutions for industrial clients. The solution shall be able to control auxiliary RAN equipment and provide bidirectional communication including but not limited to direct energy consumption readings e.g. temperature, alarms. Moreover, Smart Metering solution will collect and store power consumption, temperature and other data read out from meters installed on the site for long term data storage and near real time data analysis. The solution shall allow to monitor and proactively control auxiliary devices (e.g. active cooling, passive cooling, heating) inside the equipment shelter and analyse the data for further optimisation of energy consumption.
DATAX Sp. z o.o.
DATAX aims to complement and extend its existing network management platform RANOOK with the energy efficiency components and build a tool able to perform the course of actions that reduce energy consumption as well as provide network analysis, simulations and energy effectiveness demonstration. In general, the RANOOK platform in its present shape supports vital business processes of every mobile network operator. Addressing all of them in one platform gives the possibility to cross-correlate information coming from various sources for added value. This in return transforms into more efficient network management and cost savings.
Functionalities the extended RANOOK platform should be able to provide:
• Network simulations: the extended RANOOK platform should be able to run in simulator mode with and without connection to live network. In this mode, the tool should be capable of energy efficiency solution testing and evaluation i.e. algorithms and settings, simulations should be performed based on data acquired from the network or based on the appropriate for simulation models (traffic, load, mobility) network topology, network deployment and specific modelling requirements according to the RAN technology.
• Business case demonstration: based on the simulation results and energy consumption models, the tool should be able to present costs efficiency improvements and benefits of the applied solutions i.e. CAPEX/OPEX reduction, etc.
• Network energy efficiency monitoring: it will be an extension of the RANOOK StatView component. The StatView should enable end users to monitor network performance at the high level as well as single NE level. Besides the currently available RANOOK monitoring capabilities the extended platform should be capable of live monitoring of network energy consumption.
• Energy efficiency network optimisation: network management is a main function of the RANOOK platform which aims at network performance improvement. The energy efficiency functionality should allow optimising network parameters to reduce energy consumption and sustain network performance at the required level.
• Support for network designing process: the solution should be able to support network design process by means of network elements deployments analysis (planned and already existing), network topology assessment and parameters settings evaluation, etc., in respect of the network power consumption.
List of Websites:
Project Coordinator Office:
DATAX Sp. z o.o.
ul. Muchoborska 6, 54-424 Wrocław
Phone:+48(71) 788 58 51