accounting for the energy consumption of personal computing
TRANSCRIPT
Accounting for the Energy Consumption of Personal Computing Including Portable Devices
Pavel Somavat
Wichita State University Wichita, KS, U.S.A
Shraddha Jadhav Wichita State University
Wichita, KS, U.S.A [email protected]
Vinod Namboodiri Wichita State University
Wichita, KS, U.S.A [email protected]
ABSTRACT In light of the increased awareness of global energy consumption, questions are also being asked about the contribution of computing equipment. Though studies have documented the share of energy consumption due to these equipment over the years, these have rarely characterized the increasing share contributed by the rapidly growing portable computing segment. The portable computing device segment is widely predicted to be a dominant mode of computing and communication, and accounting for its energy consumption is necessary for energy-efficient computing in the future. This work takes a fresh and updated look at the energy consumption due to computing devices in perspective of global consumption, and pays special attention to the contribution of portable computing devices. We further quantify the impact of energy consumed by the computing sector on the environment, as well as on the electricity cost to an average residential consumer. Finally, based on the results of the study, recommendations targeted at the computer networking community are made.
General Terms Management, Measurement, Documentation.
Keywords Energy, Electricity, Computing, Portable Devices, Environment, Networking
1. INTRODUCTION The increasing role of the Internet in our lives has ushered in a tremendous growth in computing devices governed by patterns such as the famous Moore’s law. These devices are playing different roles in server farms, data centers, office equipment, among others. There is increased awareness, however, in how the world consumes energy, and its impact on the planet. It is natural to thus think about the impact of computing on global energy consumption as well. There have been many studies that document this impact [1, 2, 3].
The world is, however, changing the way it accesses the Internet, and computing in general. There is increasing relevance of portable, battery operated devices in how we handle computing and communication tasks. The first phone call over a GSM cellular phone occurred in 1991. By the end of 2007, half the world population possessed such phones. This phenomenon is
similar to the growth of computing devices in general where CPU processing power and capacity of mass storage devices doubles every 18 months. Such growth in both processing and storage capabilities fuels the production of ever more powerful portable devices. Devices with greater capabilities work with more data, and thus subsequently require greater capabilities to communicate data. This has resulted in a similar exponential growth of wireless communication data rates as well to provide adequate quality of service.
There have been no studies to date that fully document the impact of the rapidly growing segment of portable computing devices on global energy consumption, and what we can expect in the future. This requires a detailed analysis of energy consumption by these devices as compared to computing devices in general, as well as an understanding of the magnitude of these numbers in terms of global energy consumption. This accounting also needs to include the impact not only in terms of energy, but also on the impact on the environment.
In this work we profile the energy consumption of portable computing devices, and account for their impact on global energy consumption. Our contributions in this work are the following: (i) We account for global energy consumption due to all
computing devices that include server farms/datacenters, office equipment and desktops. Our statistics indicate that computing consumes more than 3% of the global electricity consumption.
(ii) We characterize the power consumption of portable devices like laptops and mobile phones, and account for their share of global energy consumption. We find that the share of energy consumed due to portable devices is 20% of the overall personal computing segment1.
(iii) We put the energy consumption due to computing devices in perspective of global electricity consumption and consider their broader impact. Our accounting indicates that emissions due to computing annually are equivalent to 4.5 million vehicles on the road. From the perspective of residential electricity consumption of an average consumer, the share due to computing is 3% in the U.S. However, when a populous, developing country like India is considered, computing consumes about 8% of the total share due to the
1 We define personal computing segment to include devices used on the front-end by consumers like laptops, mobile phones, and desktops, and devices used on the back-end like server farms/data centers, and Mobile Infrastructure. We do not include office equipment like printers, copiers etc. and network equipment like routers, switches etc.
lower electricity consumption per household compared to the U.S.
(iv) We make recommendations specific to the networking community to deal with the issue of portable device energy consumption. This is based on our results in this paper that communication/networking is a major consumer of energy in portable devices.
These contributions are significant and novel due to the fact
that we provide adequate detail about how our energy numbers are calculated for various computing sectors, and include the fast growing portable device segment. We expect our paper to allow researchers to identify the impact of individual computing sectors in terms of energy, and work towards greater energy-efficiency in future. Without such detailed accounting, the notion that computing is a significant consumer of energy will never emerge from the shadows of other segments like transportation, appliances, and lighting.
This paper is organized as follows. In Section 2 we account for energy consumption due to various sectors. In Section 3 we profile the energy consumption of individual portable computing devices. In Section 4 we estimate the total number of portable devices used globally. Subsequently, using numbers from Section 3 and 4, we calculate total energy consumption due to portable computing in Section 5. In Section 5 we further present energy consumption statistics for the personal computing segment in the context of overall global electricity consumption. In Section 6 we look at the environmental impact in terms of emissions due to the personal computing sector, and cost to consumers from a residential consumer perspective. Based on our results, in Section 7 we make specific recommendations on how to keep portable device energy consumption in check and prepare for a future of portable pervasive computing whose energy consumption cannot be ignored.
2. GENERAL ICT ACCOUNTING In this section we account for the energy consumed by devices that fall into the general class of information and computing technology (ICT). This section is intended to familiarize the reader on energy consumption trends, significant events happening in the area, projected growth, and environmental impact. We will narrow down our focus to specific energy consumption numbers in Section 4.
2.1 Data Center/Server Farm Energy Consumption and CO2 Emissions
Data centers or server farms are fast becoming a considerable source of energy consumption. Some of the factors which are driving this increase include increasing popularity of web-based applications, ever increasing number of information hungry internet users, streaming video and audio, popularity of internet based social networking and a need to store a lot of data to satisfy the user demand. Many big companies like Google, Microsoft, Yahoo, Ebay, Amazon etc. are investing very heavily into establishing newer server farms to provide support to their respective customer queries. The emerging concept of cloud computing is firing this development even further. However rapid increase in data centers implies drawing of more and more power from the grid for their functioning. This not only includes power to run the servers but also the energy costs involved in heating, ventilating, and air conditioning (HVAC), and lighting. A large
data center may require many megawatts of electricity, enough to power thousands of homes. Although not all the data center servers are power intensive, those like search engines consume a lot of power. According to Google, every query consumes about 1 KJ [4]. Estimates put the annual electricity consumption figures of Google because of these data farms up to US $ 38 million [5]. With the global computation needs going up, these farms are set to multiply thus becoming a real cause of environmental concern.
In 2006, servers and data centers accounted for an estimated 61 million MWh, 1.5% of US electricity consumption, costing about 4.5 billion dollars. According to another study by Energy Star Program, the projected power consumption by the US server farms in 2011 will reach 100 million MWh which will require about 10 new nuclear or coal powered generation plants [6]. So these farms are really becoming electricity guzzlers and a considerable source of CO2 emissions. Estimated number of servers in US is set to reach 15 million at the end of 2010, increasing from 10 million in 2005. According to Dave Douglas, Vice President, Eco-Responsibility, Sun Microsystems, the total CO2 emissions globally is 27 billion tons, where the server farms alone contribute 200 million tons of CO2 [7]. According to another study conducted by Gartner Research, Inc. all the data centers together generate 23% of all emissions produced by the ICT industry and these figures are increasing dramatically [8]. According to Environmental Protection Agency (EPA), the data centers in US produced 44.4 million metric tons (MMT) of CO2 in 2007 and the continuing trends mean that these will release approximately 79 MMT of CO2 in 2011 [6]. From the above numbers we see that the US power consumption due to server farms is set to increase from 61 million MWh to 100 million MWh in a span of 5 years. This means an annual increase of 12.7% per year. Hence the approximate power consumption due to these farms in United States can be assumed to be 84.4 million MWh. Assuming that US accounts for around half2 of the global server farms, global consumption due to server farms can be estimated to be 168.8 million MWh.
2.2 Office Equipment Energy Consumption and CO2 Emissions According to Department of Energy projections ranging from 2005 to 2015 published in 2005, office equipment will be the fastest-growing commercial end use between 2003 and 2025 [9]. Hence office equipment power consumption becomes a concern of prime importance in the fight to reduce overall energy consumption. Potential for energy savings in offices through efficient devices employing various energy efficient designs was identified in early 90s. Energy Star program was initiated in 1992 by the United States Environmental Protection Agency (EPA) as an effort to reduce the energy consumption and greenhouse gas emissions from power plants. In 1993 EPA extended the program to computers, monitors and printers. Initial motivation was to extend the energy saving features of laptops to the desktops and printers. These measures became popular partly due to governmental regulation that all the microcomputers, monitors, and printers purchased by federal agencies be Energy Star
2 It was reported in [10] that in 2005, the U.S had about 11 million servers, while the total worldwide was 26 million. Thus, in our opinion, it is reasonable approximation to assume the U.S total to be half of the global total in 2009.
compliant [11]. Started as a voluntary labeling program designed to promote energy efficient design in computer related products, this movement has extended to more than 40,000 components ranging from homes and buildings to lighting. Since its inception in 1992, Energy Star has grown from a program focusing on personal computers to a multinational program promoting over 30 products across commercial and residential markets, with thousands of program partners [12]. According to the annual savings report published in March 2009 it is estimated that the program resulted in more than $ 19 billion in total energy bill savings and at the same time reduced green house gas emissions by more than 43 million metric tons [13].
Motivated by the success of this program, various studies have been done documenting the energy consumed by office equipment with recommendations given to bring this number down; this has resulted in new start-ups with green technologies in mind [14]. A detailed study done in 2006 found that although power management modes of office computing devices is saving energy in the devices, there is still a considerable amount of energy being wasted by these devices when inactive. This study recommended PCs connected to networks for more savings potential [2]. Taking this approach even further a Japanese study explored the saving potential by shortening the power management delay time for office devices like PCs, displays, copiers, and laser printers. The results showed that such device management can save as much as 3.5 TWh per year, which is nearly equal to 2% of commercial electricity consumption in Japan [3]. So even though a lot of work has been done in the direction of energy savings for office devices, a lot can still be achieved. Therefore, further efforts from the scientific community are needed to attain more energy efficiency in offices.
2.3 Growth in Personal Computing Devices Another big issue is the rapid increase in number of computing devices globally, resulting in them becoming another major consumer of electricity. Contribution of these computing devices to the world energy consumption is becoming very significant and these cannot be ignored anymore. The data provided by Computer Industry Almanac through their annual briefings since 1993 give very detailed information for the total number of computers in use throughout the globe. Starting in 1975, the approximate number of computers in use in the world was 0.3 million out of which 0.2 million were in United States. In next five years this number rose to 4 million showing a 1233% increase. From 1980 to 1990 this increase was from 4 million to 116 million with a 2800% increase over the decade. The computer industry continued to grow at a remarkable pace in the 1990’s and the number of computers in use increased from 116 million in 1990 to 530 million in 2000 implying a 357% increase [15].
In the new millennium, this growth continued unabated although the industry reached a relative state of maturity. Estimates by Forrester Research Institute forecast the total number of computers in 2015 to hit 2 billion with the countries like India, China, Brazil and Russia accounting for more than 775 million new computers [16]. Similar estimates by Gartner believe to reach the 2 billion mark by 2014 and have forecasted the annual growth to be 12% per year [17].
Thus, from the above mentioned statistics the number of total computers in use can be estimated to be 1300 million at the end of 2009 [15]. The increase in number of computers in use since 1975 and the growth projections till 2010 are illustrated in Figure 1 based on [15, 16]. This approximation includes the desktops,
laptops, palmtops and server computers. So, summing up all the data provided by Computer Industry Almanac through its annual forecasts since 1993, the resultant estimates for the worldwide number of computers in use are depicted in Figure 1.
Fig. 1 Our projected statistics for number of computers in use worldwide
2.4 Other Related Work Significant power is consumed by mobile telecommunication infrastructure including base stations. According to a study done by mobile communications giant Ericsson in 2007, communication infrastructure contributed to approximately 0.12% of global electricity consumption and about 0.14% of global CO2 emissions [18]. So from the above data we assume that mobile infrastructure must be consuming approximately 0.15% of global electricity as per 2009. So the energy consumption is roughly 0.02 trillion KWh. As an ever increasing number of people are becoming subscribers, this will further fuel the demand for extensive coverage networks which in turn will draw more power from the grid. This will be especially significant in developing economies where current infrastructure is not enough to satisfy the increasing demand. Rising awareness about the detrimental environmental effects of increase in greenhouse gas emissions and increasing costs of energy has compelled a large number of companies and industrial segments to have second thoughts about their priorities. Therefore motivated by ethical or economical reasons a large number of companies have started giving consideration to green practices. Huge amount of savings attained by the Energy Star program also provided a positive momentum in this direction. All this has motivated companies like Nokia, Ericsson, Intel, Google, Microsoft and others to go for the innovation ensuring energy saving practices in their products. The concern for one industry is the bounty for others; in the UK alone there are around 7000 companies which are operating with environmental technologies as the sector is set to grow to $ 34 billion by the end of the decade [14]. Increasing data center energy costs and related emissions have put the whole industry in a spot of bother and large efforts are being put into place to ensure the most efficient practices. Countries like Iceland have started investing hugely in data center infrastructure to attract data centers as most of the electricity is
0
200
400
600
800
1000
1200
1400
1600
1975 1985 1989 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010Year
No.
of P
Cs (m
illio
ns)
generated from clean sources and also cold weather can be utilized for efficient cooling [19]. Besides companies, academia is also playing its part by bringing these issues to the fore. We have discussed a number of such studies earlier in this section concerning the documentation of results from the Energy Star implementation and similar other initiatives discussing various efficiency scenarios.
3 PORTABLE DEVICE ENERGY PROFILE
In this Section we take a detailed look at the power consumption profile of the two common classes of portable devices: laptops and mobile phones. We obtain the power consumption profile of these devices through a combination of our own measurements and prior work by researchers. Our measurement was based on using the Kill-A-Watt meter [20]. This meter provides the instantaneous Power consumption of any device plugged into it.
3.1 Laptops We looked at nine different laptop models belonging to multiple manufacturers and measured their power consumption. As communication is an integral part of portable computing devices, we paid special attention to the state of the wireless network interface card (WNIC). We kept the laptops in three different states: Idle with WNIC off, Idle with WNIC on (but idle), and laptop receiving streaming video through WNIC. All laptops were running windows XP or Vista, and were measured immediately on boot up. By averaging the `Idle with WNIC on’ column in Table 1, we arrive at an average consumption of a laptop around 35 W, with some approximation.
Similar numbers have been reported by prior studies with laptops [21, 22]. These studies also provided the power breakdown across different components of a laptop. We produce this breakdown here in Figure 2 for the sake of completeness of this paper. It can be seen that, for an active laptop, about 20% of the power consumption is due to the WNIC and rivals power consumption due to the Display and CPU.
Fig. 2: Power Consumption Breakdown of a Laptop actively Transmitting FTP traffic [21]
3.2 Mobile Phones Our study of mobile phone power consumption was based on only one smartphone, Samsung BlackJack, which does not have a Wi-Fi WNIC. The Bluetooth interface was kept off. The idle phone consumed about 1W while during a call it consumed 3W. Thus, during active communication, the communication interface consumes about 2/3rd the power of the device. This is confirmed by other studies done on devices with Wi-Fi interfaces like PDA’s [23]. Studies by Nokia also confirm that mobile phones typically take just over 1 Watt of power on average (1.2W to be precise).
As all emerging smartphones like the Apple iPhone and RIM Blackberry have a Wi-Fi interface in addition to the cellular interface, we believe our estimate of power consumption of mobile phones is a conservative one.
4 ACCOUNTING FOR NUMBER OF
PORTABLE DEVICES In this section we account for the number of portable computing devices used in the world. As before, we stick to only two classes: laptops and mobile phones. Based on the cumulative numbers determined in this section, coupled with the power consumed for each of these devices as presented in the previous section, we will account for global energy consumed due to portable devices annually in the following section.
4.1 Laptops Based on the total computing numbers reported in Section 2, our challenge here is to identify the number of laptops from this overall projected number of computers in world. Recent trends have shown that mobile computer segment is the most rapidly expanding segment in computer industry. Historically laptops always have cost more than their desk counterparts. With advancements in microelectronics and continuing validity of Moore’s law, laptops are becoming cheap and affordable. The advancements in Wi-Fi technology are making mobile connectivity more and more ubiquitous, and are one of the biggest factors driving the popularity of laptop computers. With ever increasing processing speed and lower costs, more and more
Model Specification Power
consumption
when WNIC off
(W)
Power
consumption
Idling with WNIC
on (W)
Power
consumption
when video
streaming through
WNIC (W)
HP pavilion dv4t 4GB RAM, 2GHz
processor 30 31 32
Dell Inspiron 1525 3GB RAM, 2GHz
processor 21 24 36
Compaq Presario
C300
2GB RAM, 2GHz
processor 28 30 34
Dell Inspiron 1440 4GB RAM, 2.2GHz
processor 18 19 24
Aspire 4730Z 2GB RAM, 2GHz
processor 27 29 31
Dell Inspiron XPS M1310
2GB RAM, 1.6GHz
processor 36 39 48
Hp pavilion dv
2000 2GB RAM, 2GHz
processor 64 71 72
Fujitsu Siemens AMILO M7440
512 MB RAM,
1.73GHz processor 29 31 35
Lenovo ThinkPad
X60 512 MB RAM, 1.6
GHz processor 21 25 28
Table 1: Laptop Power Study
pthoAqsla[lainsac
4Mtesdatopig2aw
oom(aInthathmb
5
Infocdccpd
5ASpaw1aM
people prefer laphe history of co
of laptops sold eAccording to Intquarter of 2008 surpass 50% shaaptops stood at 25]. It becomesaptops are prefencrease in futur
study we can estiaround 350 millicurrent market sh
4.2 Mobile PMobile communechnologies wh
since its inceptdecreasing priceare able to own o electricity con
phones and relagnored anymore
2008, an importachieved with worldwide [26, 2
Similarly thoriented mobile pof smart-phones million [28]. Am2006-2011) ther
ahead of Chinandonesia and Nihe US mobile m
additions from thhe above discu
mobile phones cubillion.
5 ENERGCOMPUPERSPE
n this section wfocus specificalcomputing enerdevices, we sucomputing deviconsumption. Fpercentage of thdevices alone.
5.1 Global FAccording to Summary publisprimary energy dannum from 200was estimated to17,010 Mtoe bapproximate enerMtoe which is eq
ptops these daysomputers was acexceeded those oternational Data
saw notebook are of the comp55.2% as per I
s abundantly clerred more and re as people beimate the numbeion based on nuhare and projecte
Phones nication is one hich have influetion. Thanks t
es of micro-eleca mobile phone
nsumption, we hated telecommune because of thant milestone inover 4 billion
27]. he number of mphones is on ris sold in five ymong the top re are a few surp, and almost eigeria, but the remarket tipped tohe start of 2006 ussion we can aurrently used in
GY CONSUMUTING IN GECTIVE we account for lly on the overgy used for vubsequently shices as a peFurther, we he computing se
Figures the World Enshed by Interndemand will inc
06 to 2030. In 20 be 11,730 Mtoeby 2030 [30]rgy consumptionqual to 143 trilli
s. One of the wachieved in 2008 of desktops for ta Corporation (ID
shipments intoputer market. TIDC's U.S. Quarlear from the amore and their
ecomes more mer of laptops cur
umbers presenteded trends.
of the most renced every facto the econom
ctronic componethan ever befor
ave reached a stnication infrastr
heir sheer volumn the ICT develn mobile cellu
more sophisticatese. It is estimateyears will be tri
25 growth marprises. India winequally in 3rd eal surprises staro grow by almoto the end of 20assume that thethe world to be
MED BY PEGLOBAL
global energy erall electricityvarious segmenhow the energercentage of gspecifically chegment is consu
nergy Outlooknational Energycrease at an ave006 the total ene which is projec. Using abovn in 2009 comesion KWh. Now
atershed events when the numb
the first time [24DC) survey, thi
o the U.S. markThe figure for thrterly PC Track
above studies thratio is all set
mobile. So for orrently in use to bd in Section 2 an
rapidly expandincet of human limic viability anents, more peopre. When it comtage where mobiructure cannot b
me. At the end lopment race w
ular subscriptio
ed and processined that the numbipled to reach 6rkets ranking lis the top spot, juplace are Braz
rt in 6th place wiost 66 million n011 [29]. Based oe total number approximately 4
ERSONAL
consumption any consumed. Bnts of computingy consumed bglobal electriciharacterize whumed by portab
k-2008 Executivy Agency, globerage of 1.6% pergy consumptiocted to increase
ve estimates ths out to be 12,30the next step w
in ber 4]. ird ket he
ker hat to
our be nd
ng ife nd ple
mes ile be of
was ns
ng ber 60 ist ust zil, ith net on of
4.2
nd By ng by ity hat ble
ve bal per on to he 01
will
be to mconsum(CIA) wworldwamountconsumtrillion around
5.2 CDevicNext, consumpersona
EnergyAccordmobile consumaround This fiHence mobile MWh.
numberpower assumeenergy KWh. usage oEnergyFrom omilliondesktop
eight hcomputoverall to be ar
make an approxmed globally. Aworld factbook
wide in 2008 ist and assuming
mption, the overaKWh [31]. Hen12% of global e
Computing Fces we will try t
mption by varioual computing seg
Figure 3: definitio
y consumption bding to our earl
phones currenmption for a mo
11 KWh per yegure is confirmwe can calculatdevices in the
We estimate r of laptops in consumed by a
e that a laptop isconsumption oHence overall
of all the laptopsy consumption bour assumptionsn desktop compup is about 60 W
We assume thhours a day so ter used for a energy consum
round 163.5 mill
ximation for the According to Ce
estimates, the as around 16.88a modest 2% inall number for 2nce, electricity energy requirem
Figures incl
to account forus front-end andgment as defined
Elements conon of Personalby Portable Delier estimates thntly used worldobile phone as car [32]. This com
med by our calcute the annual eneworld which co
based on our the world is aroa single laptop s used about ten
of single laptop energy consum
s would be arounby Desktop coms in previous seuters. The avera[22]. hat an average d
the energy coyear is about 1
mption because olion MWh.
total amount oentral Intelligen
amount of energy8 trillion KWh,ncrease in globa2009 comes out consumption co
ments.
uding Porta
r the approximd back-end elemd by us in Figure
sidered in ourl Computingvices:
here are aroundd-wide. Annualcited by the Nokmes out to be aroulations with smergy consumptio
omes out to be 4
earlier discussiound 350 millio
is about 35 Wn hours a day, thfor a year is ar
mption in the wnd 44.6 Million Mmputers: ection there areage power cons
desktop PC is usensumption of o175.2 KWh. Th
of desktop comp
of electricity nce Agency y consumed using this
al electricity to be 17.21
ontributes to
able
mate energy ments of the e 3.
r
d 4.2 billion l electricity kia study is ound 1.2 W.
mart-phones. on of all the 46.2 million
ion that the on. Average W [22]. We herefore the round 127.7 orld due to MWh.
e about 950 sumption of
ed for about one desktop herefore the uters comes
Energy consumption by Data Centers: As we have discussed at length, data centers are fast becoming a major source of global energy consumption. From earlier discussion, we estimate that the electricity consumption in US by server farms in 2009 will be around 84.4 million MWh. Assuming the fact that 50% of all the data centers are based in US, we can calculate the figure for world which will be 168.8 million MWh. Energy consumption for Mobile infrastructure: As per the Ericsson study, mobile communications infrastructure consumed about 0.12% of electricity globally in 2007. According to our calculations for global electricity consumption estimating 2% increase over 2008 figures, the global electricity consumption is approximately 17.21 trillion KWh in 2009 [31]. Assuming a modest increase in the infrastructure electricity consumption fraction to be 0.17%, the energy consumption due to mobile infrastructure in 2009 can be estimated to be 29.25 million MWh.
5.3 Personal Computing in Global Perspective From the above calculations, we can obtain a good approximation for total electricity consumed by personal computing devices. Combined electricity consumed by the whole computation sector for 2009 comes out to be 0.45 trillion KWh. Hence electricity consumed by computational devices is 2.61% of the global electricity consumption. Figure 4 shows the contribution of computational devices in terms of global electricity consumption.
From the mentioned numbers, we can also calculate the individual contribution of the portable devices versus complete computation segment. Overall consumption is 452.3 million MWh where as the contribution of the portable segment is 90.8 million MWh. Hence portable devices contribute 17% of the energy consumed by the computational devices. This contribution is further elaborated in Figure 5.
Fig. 4 Personal computing related electricity vs. global electricity use
As we deduced, the contribution of electricity to the overall global energy sector is about 12%. From our numbers, we can also ascertain the contribution of computation sector to the overall global energy consumption as a modest 0.31%. However considering the worsening climatic situation of the earth, this is still a significant amount and the computation sector requires more concerted efforts to reduce these numbers.
Fig. 5 Total personal computing sector electricity use vs. portable sector
6 BROADER IMPACT In this section we look at other dimensions of the figures obtained in Section 5 and earlier. We begin by presenting the energy consumption by computing devices in terms of carbon di-oxide (CO2) emissions to demonstrate environmental impact. Subsequently, we will look at the amount of residential electricity consumption by computing devices to get a consumer perspective as well.
6.1 Environmental impact One of the major points that often get unnoticed is that, although electrical energy consumption represents only 12% of the global energy requirements, its contribution to the global CO2 emissions is a whopping 40%. A study by International Energy Agency entitled CO2 emissions by sectors released in 2005 puts the CO2 emissions worldwide due to public electricity and heating to 37.2% [33]. Similar studies by European Commission’s Directorate-General for Energy and Transport released in 2009 for the year 2006 puts the CO2 emissions in the European Union (EU) zone because of energy industries and household to be around 48% [34]. Surprisingly, coal is still one of the major fuels employed to generate electricity worldwide. For example according to American Coal Foundation about 56% of electricity generated in US is done through coal [35]. Based on report on CO2 emissions by different power plants in US by Energy Information Administration, for every kWh of electricity generation, 2.117 pounds of CO2 is produced by burning coal, 1.915 pounds from petroleum and 1.314 pounds from gas [36, 37]. Hence on average electricity sources emit 1.297 lbs CO2 per kWh (0.0005883 metric tons CO2 per kWh). Thus, given that 452.3 million MWh is contributed by the personal computing segment, it responsible for about 0.27 billion tons of CO2 annually. This number is equivalent to the emissions of 4.5 million vehicles on the road when driving 10,000 miles per year considering an average car giving 21 mpg (Miles-Per-Gallon).
Another way to look at these numbers is based on the source of energy. The emissions from coal power plants is 73% of all the
Computing Electricity vs Overall Electricity Consumption
97%
3% Computing Devices
Other electricity consumption
Overall Computing Sector vs Portable Sector
17%
83%
Portable Sector
Rest Computing Sector
CgdrthgsdoAththbkegththgdeeae
6Ahnbmtoucmcthp
odaaaKa[Khc3
3
ruwnIn2oth
CO2 emitted ingenerators [38]. dioxide that haveeleased back inthat 40% of all
generation. Alsosector come soledisproportionatelother segments. Administration (he total CO2 emhis amount, ab
because of electknow that compelectricity. Hencglobal CO2 emishough the comphe global energy
global CO2 emdecrease in coefficient practiceemissions per yeattractive opporemissions.
6.2 IndividuAs per U.S. dephousehold in the number of peoplbureau [41]. Assmobile phones, ootal energy cost
using our calculconsumption duemobile phones) consumption duehe U.S would be
portable devices To get
of one of the rapdomestic energy about 56.5 Mtoeassume that the around 82.46 MKWh. The popuabout 1.14 billion42], the average
KWh. With 2 household as abcomputing would3.71%. These nu
This assumptioural areas only
with far fewer lanumbers are projndia’s middle-cl
2006 [16, 45]. Thof household comhe U.S.
nto the atmosp By burning ce been stored in to the atmospherl the CO2 emio 73% of emissely from coal whly high level of Using the estim
(EIA), we can mmissions in 2009bout 12 billion tricity generatioputing sector coe, the contributi
ssions is 0.31 bilputing sector eley consumption, t
missions. Thus, omputing sectores can prevent ear. Therefore ctunity in the gl
ual consumepartment of ener
U.S for a year ile per householduming that there
one desktop and t due to compulations in Sectioe to only portab
would be aroe to computing ie around 2.95% would be about another viewpo
pidly growing ecconsumption in
e in 2001. Baseddomestic energy
Mtoe that is rouulation of India n. Considering te household enemobile phones
bove3, the shared be 8%, while
umbers are presen
on holds typicallthe assumption
aptops and desktjected to be true lass already hadhus, for ease of mputing devices
phere by all kicoal, large qua
the coal for milre. Hence, we cassions are causions in the elechich explains wf CO2 emissionmates by the Enmake a reasonab to be 30 billiontons of CO2
on as discussedonsumes about ion of the compullion tons of CO2ectricity consumthis fragment co
achieving ever electricity coabout 5.5 mill
computing sectolobal race to co
er perspectivrgy [40], energyis about 11000 Kd is about 2.59 ae will soon be (ione laptop in a
uting would be on 5. The corrble devices (on
ound 150 KWhin an average rewhile energy co1.36%.
oint we also conconomies in the n India was 20.8 d on projected ry consumption ighly equivalentas per World
there are 4.8 peoergy consumptio, 1 laptop, ane of energy conthat of portable nted in Figure 6
ly in urban arean of 2 mobile ptops [43, 44]. Nin the very near
d at least one descomparison, we
s to be the same
ind of electriciantities of carbollions of years aan safely conclused by electricictricity generatio
why this sector hns as compared nergy Informatioble assumption n tons [39]. Out is emitted sole
d above. Also w2.61% of all thuting sector to th2. Therefore, ev
mption is 0.31% ntributes to 1%
en a modest 5onsumption wilion tons of COor presents a veontrol greenhou
ve y consumption pKWh. The averaas per U.S. censif not already) tw
typical home, tharound 325 KW
responding energne laptop and twh. Hence, energesidential home onsumption due
nsider the statistiworld: India. ThMtoe in 1990 anate of growth, win 2009 would bt to 0.959 trillioStatesmen [42]
ople per househoon would be 403d 1 desktop pnsumption duedevices would bbelow.
as in India, but phones is justifiNevertheless, ther future as most sktop or laptop be keep the numb
both in India an
ity on are de ity on
has to on of of
ely we he he en of of
5% ith O2 ery use
per ge us
wo he
Wh gy wo gy in to
ics he nd we be on is
old 38
per to be
in ed
ese of by
ber nd
like Inmuch hlower c
Figureconsumand In
7 RThe issbe alleincludeCPU’s,optimizbe impoptimizcompreeven puwe payportabl
Wmost infact thaconsumSectiondirectlycompon
7.1 PPower years fcomponThere afrom co
Thmore oshowedthe netwthe defthat pow
Hence it can ndia, the compuhigher due to gecompared to the
e 6: Compmption due tondia
RECOMMEsue of energy coviated by all m
e researchers w, network interfazations to each hroved by joint hzations focusingession, to nameut into practice
y particular attenle devices.
We believe that nn reducing portaat an active netw
mes the most pn 3. Our folloy at the networknent approaches
ower Manamanagement ha
for computing dnents have beenare at least two aontinued optimizhe first area is often. An informd that most werework card, and ufault settings of wer managemen
be observed thauting device eneeneral householdU.S.
parison of o personal com
ENDATIONonsumption of c
members of the working on diffefaces, memory, thardware compohardware optimig on operating s a few, would over the last de
ntion on how to r
networking reseable device enerwork interface i
power among aowing recommeking communitys.
agement as been an activdevices as well n optimized to areas where furtzations of compoto make users
mal survey of 50e unaware of poused the default laptops studied
nt was disabled b
at in developingergy consumpti
d energy consum
household mputing devic
NS computing deviccomputing com
erent aspects likto talk of a few
onent would helpizations. Similarsystems, data cabe helpful as s
ecade or so. In treduce energy co
earcher’s can corgy consumptionis likely to be tll components, endations are y, including pos
ve area of researas appliances [lower power co
ther work can beonents individuautilize power m
0 students at ouower managemen
settings. A simpin Section 3 ear
by default.
g economies on share is
mption being
electricity
ces in U.S.
ces can best mmunity that
ke displays, w. Individual p, and could rly, software aching, data studied, and this section, onsumed by
ontribute the n due to the the task that
as seen in thus aimed ssible cross-
rch over the 12]. Device onsumption. e done apart ally. management ur university nt modes of ple check of rlier showed
The second area, once users are aware of power management options, is to enable tunable software based trade-offs between energy and performance. For example, wireless cards can be kept in sleep mode longer at the expense of higher delay of received packets, for say, a VoIP application. Or, the resolution of a video being streamed could be reduced for reduced energy consumption. The creation of a general framework across different computing platforms to allow such tunable tradeoffs is sorely required. If the hardware equivalent of a ‘knob’ could be developed that allows each users to tune it to their own tolerable performance levels, it would make power management more acceptable.
7.2 Battery Management When batteries are necessary, as is the case when a user is mobile, better usage of batteries can get better efficiencies. Protocols and Algorithms should be designed to provide time for battery charge recovery, and use less current draw when possible. For example, medium access control (or even upper layer) protocols could be designed that sends and receives packets in short bursts, that allows nodes enough time in sleep mode in between bursts [46]. Greater utilization of battery capacity can reduce power consumed of the electricity grid.
7.3 Adjust Optimization Metrics Protocols and algorithms should be designed keeping total energy consumption as a metric, as opposed to individual node lifetimes when considering a collaborative network in specific scenarios. For example, consider the case of a static wireless mesh network. It may be possible to replace the batteries in such nodes periodically. In such an instance, protocols must be designed to minimize total energy consumption of the network as opposed to each node trying to minimize its energy consumption [47]. Another example is that of an ad hoc network in a conference room where a power outlet for each node is easily available and all nodes are plugged in. Again, in this scenario, the total power consumption by all nodes should be minimized as opposed to the common metric of individual node consumption.
7.4 Utilize Energy Harvesting Recently, there are many products on the market which allow the use of energy harvesting techniques to power portable devices. For example, the Solaris product (cost about $360) from Brunton can charge laptops, while the SolarPort product (cost $120) from the same manufacturer can charge smaller devices like phones. These devices can become more main-stream with time as prices drop due to scale, or through government subsidies to encourage more renewable energy use. From a networking researcher’s perspective, there is a need to understand how to design protocols to utilize such sources of energy effectively. For example, during periods where plenty of energy is available, portable devices could do more of delay-tolerant communications. Once the energy harvest is unavailable, the device can ramp down to minimal operating modes similar to the power management paradigm.
8 CONCLUSIONS We presented a study that accounted for energy consumed by various computing devices, including the growing portable device segment. Our accounting shows that computing consumes about 3% of the global electricity consumption. Of this, about 20% is
contributed by portable devices, a significantly high share that was unaccounted for in previous work. This statistic should serve as a reminder to researchers that energy should also be treated as a resource consumption issue in portable computing devices, as opposed to the traditional operating lifetime metric. Our study also shows that the CO2 emissions due to the computing sector are much higher due to the fuel source used for power, and efficiencies in the computing sector can make a bigger impact on global CO2 emissions than efficient practices in other sectors. From an individual consumer perspective, we also demonstrate that the cost of electricity for computing rivals that of many other common household appliances, and requires careful thought on usage behavior. This is especially true in developing countries like India where average household energy consumption is lower than the U.S., and results in the share of computing devices being higher.
REFERENCES [1] G. Fettweis and E. Zimmermann.ICT Energy Consumption - Trends and Challenges in Proceedings of the 11th International Symposium on Wireless Personal Multimedia Communications (WPMC'08), Lapland, Finland, 08. - 11. September 2008. [2] Carrie A. Webber, Judy A. Roberson, Marla C. McWhinney, Richard E. Brown, Margaret J. Pinckard, John F. Busch, After-hours power status of office equipment in the USA, Energy 31 (2006) 2487–2502 [3] Kaoru Kawamoto, Yoshiyuki Shimoda, Minoru Mizuno, Energy saving potential of office equipment power management, Energy and Buildings 36 (2004) 915–923 [4] U. Hölzle, “Powering a Google Search,” Official Google Blog, Jan. 2009. [5] A Qureshi, R Weber et al. Cutting the Electric Bill for Internet-Scale Systems, SIGCOMM’09, August 17–21, 2009, Barcelona, Spain. [6] Report to Congress on Server and Data Center Energy Efficiency Public Law 109-431,” EPA ENERGY STAR Program, August 2, 2007, [7] Data Center Energy Savings, available at http://perfectsearchcorp.com/data-center-energy-savings [8] Gartner: Data Centres Account for 23% of Global ICT CO2 Emissions,” November 5, 2007, available online at http://engineers.ihs.com/news/gartner-datacentre-co2.htm [9] US Department of Energy. Annual energy outlook 2005 with projections to 2025. Washington, DC: Energy Information Administration, DOE/EIA-0383 (2005), 2005 [10] Jonathan G. Koomey, “Estimating total power consumption by servers in U.S and the world”, Technical Report, February, 2007. Available at: www.koomey.com. [11] C.A. Webber, R.E. Brown & J.G. Koomey, Savings estimates for the ENERGY STAR voluntary labeling program, Energy Policy 28 (2000) 1137}1149 [12] R. Brown, C. Webber, J.G. Koomey, Status and future directions of the Energy Star program, Energy 27 (2002) 505–520 [13] ENERGY STAR® Overview of 2008 Achievements, http://www.energystar.gov/ia/partners/publications/pubdocs/2008%204%20pager%203-12-09.pdf [14] The green shoots of hi-tech success, report available at http://news.bbc.co.uk/2/hi/technology/6983117.stm [15] Annual Forecast Statements 1993-2009, Computer Industry Almanac Inc., available at http://www.c-i-a.com/ [16]Press Release, Forrester Research Inc., available at http://www.forrester.com/ER/Press/Release/0,1769,1151,00.html
[17] Gartner Says More than 1 Billion PCs In Use Worldwide and Headed to 2 Billion Units by 2014, Press Release, available at http://www.gartner.com/it/page.jsp?id=703807 [18] Sustainable energy use in mobile communications (August, 2007). White paper by Ericsson [19] Iceland looks to serve the world, report available at http://news.bbc.co.uk/2/hi/programmes/click_online/8297237.stm [20] Kill-A-Watt Meter, online available at http://www.p3international.com/products/special/P4400/P4400-CE.html [21] Aqeel Mahesri and Vibhore Vardhan, Power Consumption Breakdown on a Modern Laptop, Workshop on Power Aware Computing Systems, 37th International Symposium on Microarchitecture (PACS), December 2004. [22] Intel corporation, PC Energy-Efficiency Trends and Technologies, online available at cache-www.intel.com/cd/00/00/10/27/102727_ar024103.pdf [23] Vijay Raghunathan, Trevor Pering, Roy Want, Alex Nguyen, Peter Jensen, "Experience With A Low Power Wireless Mobile Computing Platform," islped, pp.363-368, Proceedings of the 2004 International Symposium on Low Power Electronics and Design (ISLPED'04), 2004 [24] Notebook PC Shipments Exceed Desktops for First Time in Q3, iSuppli Applied Market Intelligence, Press Report, available at http://www.isuppli.com/News/Pages/Notebook-PC-Shipments-Exceed-Desktops-for-First-Time-in-Q3.aspx? [25] Notebook Shipments Surpass Desktops in the U.S. Market for the First Time, IDC Press Release, available at http://www.idc.com/getdoc.jsp?containerId=prUS2149320 [26] Measuring the Information Society, “The ICT Development Index”, International Telecommunications Union Society-2009. [27] Walsh, Mark. 2009. “US Mobile Broadband Users to Surpass 140 Million by 2013.” January 6, available at http://www.mediapost.com/publications/ [28] Half the World Will Use a Mobile Phone by 2009, available at http://www.cellular-news.com/story/15674.php [29] Half world's population 'will have mobile phone by end of year' online news article available at http://www.guardian.co.uk/technology/2008/sep/26/mobilephones.unitednations [30] World Energy Outlook 2008, Executive Summary, International Energy Agency, available online at http://www.iea.org/weo/docs/weo2008/WEO2008_es_English.pdf [31]CIA World Factbook, Annual Report, available at http://www.indexmundi.com/world/electricity_consumption.html
[32] Report by Nokia online available at http://eu.europa.eu/environment/ipp/pdf/nokia_mobile_05_04.pdf [33] Carbon Dioxide Emissions by Economic Sector 2005, Report by International Energy Agency [34] CO2 Emissions by Sector, EU ENERGY IN FIGURES 2009, Report by Directorate-General for Energy and Transport, European Commission [35]FAQs about coal, American Coal Foundation available at http://www.teachcoal.org/aboutcoal/articles/faqs.html [36] Carbon Dioxide Emissions from the Generation of Electric Power in the United States, report available at http://www.eia.doe.gov/cneaf/electricity/page/co2_report/co2emiss.pdf [37] Report from Carbonfund available online at http://www.carbonfund.org/site/pages/carbon_calculators/category/Assumptions [38]Power Scorecard, Electricity from coal available online at http://www.powerscorecard.org/tech_detail.cfm?resource_id=2 [39]World CO2 Emissions to Rise 75 Pct by 2030 – EIA, at http://www.planetark.com/dailynewsstory.cfm/newsid/36924/story.htm [40] Energy savers, U.S. department of energy. Retrieved from http://www1.eere.energy.gov/consumer/tips/appliances.html [41]Quick facts, U.S. census bureau, retrieved from, http://quickfacts.census.gov/qfd/states/00000.html [42] World Population Trends, World Statesmen Forum, retrieved from, http://www.worldstatesmen.org/India.htm [43] Average Residential Energy Consumption per Household, retrieved from http://www.nfhsindia.org/ on October 23, 2009 [44] Telecom Regulatory Authority of India (TRAI) Annual Report available online at http://www.trai.gov.in/annualreport/TRAIAR2008-09E.pdf [45] Wired Generation, India Today, available online at http://www.india-today.com/itoday/20061120/cover.html [46] C. F. Chiasserini , R. R. Rao, Pulsed battery discharge in communication devices, Proceedings of the 5th annual ACM/IEEE international conference on Mobile computing and networking, p.88-95, August 15-19, 1999, Seattle, Washington, United States [47] Vinod Namboodiri, Lixin Gao, Ramakrishna Janaswamy, "Power Efficient Topology Control for Wireless Networks with Switched Beam Directional Antennas", Elsevier Ad Hoc Networks, Vol. 6, No. 2, April 2008.