Greening information and communication technologies

21/12/2016
OAI : oai:www.see.asso.fr:1301:2016-5:17786
contenu protégé  Document accessible sous conditions - vous devez vous connecter ou vous enregistrer pour accéder à ou acquérir ce document.
Prix : 10,00 € TVA 20,0% comprise (8,33 € hors TVA) - Accès libre pour les ayants-droit
 

Résumé

Greening information  and communication technologies

Métriques

7
4
401.52 Ko
 application/pdf
bitcache://4381999695ebdb9163d1345ef3b9ef5c86d96517

Licence

Creative Commons Aucune (Tous droits réservés)
<resource  xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
                xmlns="http://datacite.org/schema/kernel-4"
                xsi:schemaLocation="http://datacite.org/schema/kernel-4 http://schema.datacite.org/meta/kernel-4/metadata.xsd">
        <identifier identifierType="DOI">10.23723/1301:2016-5/17786</identifier><creators><creator><creatorName>Anne-Cécile Orgerie</creatorName></creator></creators><titles>
            <title>Greening information  and communication technologies</title></titles>
        <publisher>SEE</publisher>
        <publicationYear>2016</publicationYear>
        <resourceType resourceTypeGeneral="Text">Text</resourceType><dates>
	    <date dateType="Created">Wed 21 Dec 2016</date>
	    <date dateType="Updated">Thu 26 Jan 2017</date>
            <date dateType="Submitted">Fri 20 Apr 2018</date>
	</dates>
        <alternateIdentifiers>
	    <alternateIdentifier alternateIdentifierType="bitstream">4381999695ebdb9163d1345ef3b9ef5c86d96517</alternateIdentifier>
	</alternateIdentifiers>
        <formats>
	    <format>application/pdf</format>
	</formats>
	<version>30416</version>
        <descriptions>
            <description descriptionType="Abstract"></description>
        </descriptions>
    </resource>
.

URSI FRANCE 2016DOSSIER 2 90 REE N°5/2016 Greening information and communication technologies Par Anne-Cécile Orgerie Chargée de recherche, docteur en informatique, CNRS, IRISA Rennes Les technologies de l'information et de la communication (TIC) sont devenues indispensables dans notre vie de tous les jours. Le nombre d'équipements connectés augmente rapidement, tout comme leur capacité. Cependant, cette croissance entraîne une consommation électrique inquiétante. L'informatique verte vise à réduire les impacts des TIC sur l'environnement. Comment quantifier les impacts des TIC ? Quelles sont, dans l’état actuel des connaissances, les techniques principales permettant de construire une informatique durable ? ABSTRACT Energy and information and communica- tion technologies Information and communications technologies (ICT) are often cited as a major lever to reduce the energy consump- tion and greenhouse gas emissions of other industry fields. They are indeed able to optimize and automate industrial processes, but also to offer new innovative solutions. For ins- tance, the usage of ICT can limit transportation – which is a huge greenhouse gas emitter – through videoconferencing, reduce human travels and, through car-sharing websites, optimize them. From online journals to smart building, ICT is progressively invading most of industries, often with an eco- logical argument. However, ICT themselves consume substantial amounts of electricity and are responsible for weighty greenhouse gas emissions. Consequently, exploiting ICT capabilities to come to the rescue of other industries’ sustainability leads to rebound effects that can lower – or even reverse – their benefits. The complexity of the connections between ICT and sustainability comes from the fact that ICT devices can be both actors and consumers at the same time, and thus, they also need to turn green. In this article, we explore this complex relation by first looking at the current energy consumption of ICT. Then, we identify several state-of-the-art approaches for greening ICT. Finally, we conclude by presen- ting the arduous challenge of having an overall view of the potential levers enabling a greener ICT. Current situation: ICT energy consumption According to the abundant literature on the subject, ICT industry would account for 2 to 10 % of global CO2 emissions. The various studies dealing with this complex evaluation do not agree on an exact number to measure ICT impact. As a comparison, aviation is estimated to represent 2 % of global CO2 emissions. However, one can notice that, firstly, aviation growth is slower than the increase of ICT industry, and second- ly, aviation has never been considered as a “green” industry. The same complexity can be observed while looking for an evaluation of the global energy consumption of ICT. The consensus is around 5% according to the French Academy of Technologies (1). Currently, for this global energy consumption, the following breakdown is observed: approximately one third for data centers, one third for communication networks and one third for user equipment (computers, monitors, etc.) (Ibid.). Each portion is complex to compute due to the rapid adoption of new ICT devices, the blurred frontiers between these three categories, and the diversity of technologies used in ICT, which do not have identical impacts in terms of electricity consump- tion – wireless and wired telecommunication technologies for instance. As an example, in 2013, Coroama and Hilty (2) reviewed the literature to assess Internet energy intensity, and found that: “Estimates published over the last decade diverge by up to four orders of magnitude — from 0.0064 kilowatt-hours per gigabyte (kWh/GB) to 136 kWh/GB”. Around 2008, however, the emergence of cloud computing promised a massive mutualization of servers through isolation and co-location of multiple virtual machines on the same phy- sical server. In addition, advances in multicore architectures have improved the energy efficiency of servers: for the same calculation, less energy is needed. However, ICT followed the Jevons paradox (rebound effect), formalized in 1865 about coal: increasing the efficiency with which a resource is used leads to an increase of its total consumption because, in the same time, this improved efficiency reduces its relative cost, thus accelerating its economic growth. Indeed, in spite of significant improvements in its energy efficiency, the energy consumption of ICT has continued to increase from 3.3 % to 4.7 % of global consumption between 2007 and 2012 (1). REE N°5/2016 91 Greening information and communication technologies For the end-user, it is difficult to grasp the energy consumption of the virtual Cloud resources he is using: these resources seem invisible and are often free of charge (Facebook, Twitter, Gmail, etc.). However, the information appetite of Big Data, big networks and big infrastructure una- voidably leads to big power (Mills, 2013). In 2012, it was esti- mated that, each day, 145 billion emails were exchanged, 4.5 billion searches were launched on Google, 104,000 hours of video were uploaded on Youtube and 40 000 gigabytes of data were produced at the Large Hadron Collider (3). Such widely used Cloud services are supported by gigan- tic infrastructures. As an example, for 2010, Google used 900,000 servers and consumed 1.9 billion kWh of electri- city (4), and the power demand of a single datacenter can reach up to 100 MW. In 2012, the number of data centers worldwide was estimated at 509,147, consuming roughly the output of 30 nuclear power plants (5). On the other side, it is estimated that in 2017, Internet users will rely on an average of five connected devices per person, with more than 20 bil- lion devices connected. All portions of ICT technologies seem concerned by this inexorable energy consumption growth. The urgency of the situation calls for significant improvements and changes, in particular in user habits. While this last point appears arduous, research in green computing has delivered interesting pathways to make ICT more sustainable. Towards greening ICT Energy has always been a matter of concern in certain domains of ICT, like sensor networks and battery-constrained devices. However, for other ICT systems, like data centers, it has only recently become an issue (6). The Gartner’s annual hype cycle for emerging technologies shows this phenome- non: green IT first appeared in 2008, directly at the peak of the wave; in 2009 it was on the declining side of the wave, and surprisingly by 2010, it had disappeared. Reducing ener- gy consumption of ICT devices is a challenging issue that should be addressed at different levels: hardware, software and user levels. Improving energy efficiency In order to improve energy-efficiency, it is necessary to identify the biggest consumption items for each ICT field. For instance, in data centers, cooling systems are typically estimated to consume 33 % of the total value (7). A green evolution consists in using free cooling: using outside air to cool servers. This technique leads companies to locate their facilities in regions and countries with cold climate, such as Sweden, Finland or Iceland. Energy savings can be made at all hardware levels: from processor chips to data center infrastructures. The energy- efficient design of low-power processors or the use of dedi- cated hardware (e.g. Graphics Processing Unit) can reduce the energy cost per operation1 . Energy efficiency indicators were defined to characterize data centers. The most common is the PUE (Power Usage Effectiveness) introduced by the Green Grid consortium in 2006. It is determined by the ratio of the total energy consumed by the center and the part consumed only by the “useful” facilities: server, storage and network. However, this 1 NDLR : sur la question des data centers, le lecteur pourra se reporter à la REE 2015-4, pp. 54-65 – Les data centers – Jean-Pierre Hauet. Figure 1: Salle abritant les serveurs et le stockage informatique du laboratoire d’Annecy-le-Vieux de physique des particules (LAPP) et du mésocentre MUST (Méso infrastructure de calcul et de stockage) – Source : ©CNRS Photothèque. URSI FRANCE 2016DOSSIER 2 92 REE N°5/2016 indicator has two major drawbacks: it does not account for the actual use of computing and storage resources and does not take into consideration the origin of the energy consu- med. Large data centers manufacturers no longer hesitate to display their PUE and make its decline a commercial argu- ment. For instance, Google shows that its average PUE has decreased from 1.21 in 2009 to 1.12 in 2016. Green computing has also benefitted from several stan- dards and regulations. Tangible efforts have been made on the energy consumption of ICT equipment (EnergyStar stan- dard for reduced consumption in standby and off appliances, 80plus for energy-efficient power supplies) and the restric- tions about the use of hazardous substances (European di- rectives RoHS, REACH). Other works, not from the legislative side but rather on recommendations, should also be men- tioned, like the European Code of Conduct on Data Centers advising infrastructure admins to have greener habits, by increasing the data center room temperature for instance. It seems that only regulations and norms can fight against certain bad manufacturing practices, in particular those leading to obsolescence: e.g. when a device is replaced although it is still in good working state. This happens when the device is in excess due to software evolution. For example, the end of Windows XP support is effective since 2014. This means that computers with this operating system will become potentially vulnerable to new securi- ty vulnerabilities and cannot be used without safety risks. According to a survey conducted in March 2016, 11.24 % of computers worldwide, roughly 200 million computers, are still using Windows XP (8). Increasing resource utilization In addition to improve energy-efficiency, it is also requi- red to increase resource utilization in order to reduce energy waste. Indeed, unused but powered-on resources consume non-negligible amounts of electricity. In 2010, the Green Grid consortium carried out a survey about unused servers in 188 data centers, mostly located in the United States (9). They estimate that on average 10 % of the servers are never used, hence wasting energy. Another study conducted in 2008 shows that data cen- ter server utilization rarely exceeds 6 % (10). As the energy consumption of servers is not proportional to their workload they consume large amounts of energy even when not in use but simply powered-on – 40 and 60 % of their maximum consumption when fully loaded (6). It is therefore vital that these servers have standby capabilities (more energy-effi- cient mode) and remote wake-up mechanisms, in order to adapt the computing capacity on user demand. Two approaches are commonly employed to reduce this energy waste: sleeping techniques and slow down techniques (6). The first one consists in putting to sleep or standby mode unused resources, and in using remote wake-up mechanisms such as IPMI (Intelligent Platform Management Interface). This is a standard hardware that operates independently from the operating system and allows administrators to manage a system remotely through a direct connection. This interface can also be used to switch nodes on and off remotely. The slow down technique for processors is called DVFS (Dynamic Voltage and Frequency Scaling) and gives proces- sors the ability to adjust their working frequency and power consumption to save energy (Ibid.). It is then necessary to implement smart algorithms to decide when and how to switch between the available processor frequencies depen- ding on the workload. Software also does not escape to the ever-growing ener- gy needs; needs that are increased by the multiplication of available applications. In 2014, a survey shows that smart- phone owners have an average of 35 installed applications, of which only 11 are used every week and 12 are never used, and this number has increased since the previous study by Harris Interactive (11). However, all these installed applica- tions consume computing and storage resources. Indeed, an application that is never used can be programmed to send periodic notifications to remote servers (e.g. location data) and thus have significant energy consumption. Software eco-design consists in optimizing their environ- mental footprint and allows real energy gains. It relies on the use of more energy-efficient algorithms, reducing the exe- cution time (and thus CPU utilization), developing precise specifications to eliminate unnecessary features and optimi- zation of generated data volumes. For example, a modifica- tion of the mobile version of the Wikipedia website scripts reduces by 30 % the energy needed to download and render its webpages. One should go back to software practices of the past time when code was optimized because of the high cost of resources available on the market (RAM, processor speed). Increasing energy awareness of users Unlike in the case of data centers, the user can act directly on the software and devices he uses, and has therefore a key role to play in reducing the ICT impact on sustainability. However, this requires him to be informed of the available levers he has access to, and the actual consumption of his applications. Some software tools provide real time measu- rements or estimations of the energy consumption of pro- cessors, such as the Intel Power Gadget (12). It would be REE N°5/2016 93 Greening information and communication technologies interesting to extend its functionalities, so that users can esti- mate the energy consumption of each of their applications including all resources used (CPU, memory, network card, hard drives, etc.). The PowerTOP software (for Linux) goes even further in this direction (15): it allows a diagnosis of the energy consumption of a computer; it provides configuration tips to reduce consumption and offers an interactive mode in which the user can test different configurations of his system. Such software can provoke a real awareness of users regarding their impacts associated with ICT on electricity consumption. As such, the approach proposed by this software would de- serve to be extended to other energy-consuming electronic devices (monitors, smartphones, etc.) and could thus pro- mote user empowerment. Increasing energy-awareness of users could also allow ser- vice providers to save energy by offering different options to users. For instance, an user can accept to wait a given amount of time for its service to be executed if he is informed that waiting is going to save a certain amount of energy because his request will be consolidated with other requests. Such a system, offering different options to the user and displaying a clear trade-off between energy saving and performance degradation, can save significant amount of energy. A user could also be willing to delay his request processing in order to benefit from renewable energy sources. In the case of specific collaborative applications, a user could accept to host on his own device part of the compu- ting system for processing requests from his neighbors and thus, save energy in the telecommunication networks nee- ded to connect each user to the service provider. However, in all these techniques, energy-awareness of the users may be insufficient for adopting energy-aware behavior and adequate incentives (financial or motivational) need to be proposed to ensure the success of such approaches. For instance, a car- bon tax may encourage users to wait for renewable energy availability, or an added social value may incite users to buy greener services or products like Fairphone – a smartphone designed and produced with high sustainability criteria. The whole loop challenge Finally, we think that the main issue in greening ICT resides in the whole loop challenge, that is to take into account all the parameters required to ensure substantial energy savings. For instance, when an user submits a Google request from a connected device (such as a search for instance), it is routed to a data center that processes it, computes the answer and sends it back to the user. Google owns several data enters spread across the world. For performance reasons, the cen- ter answering the user’s request is more likely to be the one closest to the user. However, this data center may be supplied by a far away energy source. A more distant data center may Figure 2: Cœur du réseau local de l’IRISA : commutateur/routeur/IP. 40 ports Gigabit Ethernet fibre optique – Source :CNRS. URSI FRANCE 2016DOSSIER 2 94 REE N°5/2016 be using a renewable energy source, which happens to be close by and available at the time of the request. This request might have consumed less energy, or a different kind of energy (renewable or not), if it had been sent to the second data cen- ter. In this case, the response time would have been increased but maybe not noticeably: a different trade-off between per- formance and energy-efficiency could have been adopted. On the other side, considering only the usage phase of the ICT equipment is biased, as other life-cycle phases can be more consuming. For instance, a use phase of five years of a typical server can constitute only half of the CO2 emissions of its entire life cycle: its manufacturing and its distribution completes the other half (14). Concerning the impact on greenhouse gas emission, life-cycle assessment shows that for servers, the use phase is predominant. However, in terms of user devices (computers, tablets, smartphones, etc.), the manufacturing phase is significantly more impactful than the use phase in France (15). One can wonder if we are going the good way: ancient telephone receivers were only connected to the telephone network. They did not require to be connected to the power grid, while current smartphones require their batteries to be recharged every few days. Smartphones are indeed offe- ring new functionalities compared to old phones, but one can wonder if these new functionalities, which create new practices, do not also generate new needs, a multiplication of devices with an important capability overlapping between these devices, and in the end, an alarming and unnecessary increase in global energy consumption. Moreover, this pheno- menon is racing fast: we exceeded 1.4 billion smartphones sold in one year in 2015, an increase of almost 10 % compa- red to 2014 (16) and, in 2015, the average life duration of first- hand smartphones before replacement fell under two years. In order to reduce ICT energy consumption, we could limit its utilization. However, the enforcement of such an idea would probably require strict international regulations. Currently, to our mind, raising the energy-awareness of users seems to be the only sustainable way of greening ICT. References [1] Académie française des technologies, “Impact des TIC sur la consommation d’énergie à travers le monde”, EDP Sciences, report 2015. [2] Vlad Coroama and Lorenz Hilty, “Assessing Internet energy intensity: a review of methods and results,” Environmental Impact Assessment Review, vol. 45, pp. 63-68, 2014. [3] Denis Delbecq, and Grégory Fléchet Fabrice Demarthon, L'AUTEUR Anne-Cécile Orgerie est chargée de recherche au CNRS à l’IRISA à Rennes. Elle a reçu sa thèse en informatique de l’Ecole normale supérieure de Lyon en 2011. Ses travaux de thèse ont été récompensés par le prix de la recherche en système attribué par le chapitre français de l’ACM SIGOPS en 2013 et le prix jeune chercheur de la ville de Lyon en 2014. Ses travaux portent sur l’efficacité énergétique dans les sys- tèmes distribués et les réseaux. REE N°5/2016 95 Greening information and communication technologies “The Big Data Revolution”, CNRS International Magazine, vol. 28, pp. 20-23, Jan. 2013. [4] Jonathan Koomey, “Growth in Data Center Electricity Use 2005 to 2010”, Analytics Press , 2011. [5] James Glanz, “Power, Pollution and the Internet”, The New- York Times, Sept. 2012. [6] Anne-Cécile Orgerie, Marcos Dias de Assunção, and Laurent Lefèvre, “A Survey on Techniques for Improving the Energy Efficiency of Large Scale Distributed Systems”, ACM Computing Surveys, vol. 46, pp. 1-31, Apr. 2014. [7] Albert Greenberg, James Hamilton, David Maltz, and Parveen Patel, “The Cost of a Cloud: Research Problems in Data Center Networks”, ACM SIGCOMM Computer Communication Review, vol. 39, pp. 68-73, 2008. [8] Net Market Share. (2016) Desktop Operating System Market Share. [Online]. HYPERLINK “https://www. netmarketshare.com/operating-system-market-share. aspx%20” https://www.netmarketshare.com/operating- system-market-share.aspx [9] Mark Blackburn and Andy Hawkins, “Unused Servers Survey Results Analysis”, The Green Grid, 2010. [10] James Kaplan, William Forrest, and Noah Kindler, “Revolutionizing Data Center Energy Efficiency”, McKinsey & Company, 2008. [11] Harris Interactive, “Usages & attitudes vis-à-vis des applications mobiles”, survey, 2014. [12] Intel. (2014) Intel Power Gadget, software. [Online]. HYPERLINK “https://software.intel.com/en-us/articles/intel- power-gadget-20” https://software.intel.com/en-us/articles/ intel-power-gadget-20 [13] Powertop. (2015) Powertop software. [Online]. HYPERLINK “https://01.org/powertop” https://01.org/powertop [14] Fujitsu, “Fact Sheet: CO2-Footprint of a Desktop-PC ESPRIMO E9900”, Fujitsu, 2010. [15] Françoise Berthoud, Eric Drezet, Laurent Lefèvre, et Anne-Cécile Orgerie. (2015) « Sciences du numérique et développement durable : des liens complexes ». [Online]. HYPERLINK “https://interstices.info/jcms/p_84005/ sciences-du-numerique-et-developpement-durable-des- liens-complexes” https://interstices.info/jcms/p_84005/ sciences-du-numerique-et-developpement-durable-des- liens-complexes [16] Anthony Scarsella and William Stofega, “Worldwide Smart- phone Forecast Update, 2015-2019”, IDC report, 2015. [17] Kris De Decker, “Why We Need a Speed Limit for the Internet”, Low-Tech Magazine, 2015.