1

THE DARK SIDE OF THE TUNE: THE HIDDEN ENERGY COST OF DIGITAL MUSIC CONSUMPTION

2 3

Foreword – Keith Harris, Chair, MusicTank 4

DIGITAL CONTENT – WHAT’S THE ISSUE? Introduction 5Commentary: Iain Hossack, Sustainability Consultant 6Assumptions 8Understanding storage capacity 9

1. DATA TRAFFIC AND ENERGY CONSUMPTION1.1 Scale and composition 101.2 Analysis 111.3 Digital vs physical 12

2. DIGITAL CONTENT2.1 Unlicensed file sharing 152.2 Downloading and streaming 16

3. TECHNOLOGY3.1 Wireless vs terrestrial 17Commentary: Ralph Simon, Chairman Emeritus, MEF 223.2 Mobile communication 243.3 Global mobile internet penetration 24

4. SUSTAINABILITY4.1 Sustainability for the rights holders 254.2 Energy efficient distribution 274.3 A solution for the future? 28

5. CONCLUSIONS 31 APPENDIX 1 Background 32APPENDIX 2 Data centres, data networks and energy consumption 33APPENDIX 3 Further reading 44

ACKNOWLEDGMENTS 45ABOUT THE AUTHOR 46ABOUT MUSICTANK 47REPORT PARTNER 49

CONTENTS

A MUSICTANK REPORT

by DAGFINN BACHR&D Director, Bach Technology ASOctober 2012

4 5

In 2007, when YouTube was barely two-years old, and well before the widespread adoption of cloud computing, Gartner, a respected market research company, estimated that the computing industry was responsible for two per cent of global CO2 emissions, 23 per cent of which were attributable to data centres on which the Internet depends.

For some reason this issue seems to have flown under the radar of many of those making decisions about the future of the delivery of media content to our desktops, mobiles and other devices.

Dagfinn Bach (Bach Technology) brought this subject to MusicTank’s attention following a study addressing the uncontrolled growth in data traffic, as part of an extensive R&D project on the future of peer-to-peer with partners including the Fraunhofer Institute and University of Bergen.

With the increasing popularity of licensed streaming services, where music and other creative content is ‘accessed’ rather than ‘owned’, this report is intended to be a springboard for a wider debate about whether we are giving sufficient thought to the energy consumption of digital music consumption in the search for successful new business models.

Keith Harris, Chair, MusicTankSeptember 2012

FOREWORD DIGITAL CONTENT - WHAT’S THE ISSUE?

The recordings business is seeing growing returns from a burgeoning number of digital distribution channels encompassing streaming, downloading and cloud storage models.

As well as making music much easier and quicker to access, an expected by-product of this growth has always been a decrease in the perceived heavy environmental cost associated with physical products, including production and pressing of CDs/vinyl, shipping, storage and warehousing.

Yet there is a hidden cost to digital that receives barely a mention outside the ICT (Information, Communication and Technology) sector – energy.

Digital music isn’t distributed in an environmental vacuum. While pressing plants for CDs and vinyl are becoming rarer, and with fewer lorries on the road transporting stock to stores, the growth in data traffic caused by digital content services comes with its own environmental risks and problems.

The embedding and signposting of rich, high bit-rate files required by various digital content services – especially online video – depends on sprawling server farms and a complex, energy-sapping network infrastructure. This is on top of the energy consumed in device manufacture and operation of a vast array of devices.

From this perspective, do ever more complex cloud, mobile and streaming services represent sustainable consumption models or do they present us with an environmentally unsustainable digital future?

While this paper indicates what the carbon footprint implications are, where the biggest problems lie and what might be done to reduce energy consumption, it is only intended to be a starting point for a wider and much-needed conversation.

As such the paper is a clarion call for further urgent research concerning the environmental impacts of burgeoning digital content industries.

INTRODUCTION

6 7

This paper considers some blue sky solutions to online content delivery which, inter alia, offer significantly reduced energy usag. Any solution that complements global energy reduction activities with their associated costs, can only be welcomed; at its basic level, reduced costs increase margins.

Notwithstanding this, a word of caution is offered for those about to enter the debate. Despite best efforts by governments, lawyers and academics, nothing can escape the principles of supply and demand, tending towards coalescence around convenience with price.

Anything we propose or legislate for requires alignment with the needs, aspirations and convenience of both businesses and consumers.

Iain HossackSeptember 2012

USER LISTENING INTENSITY

Is the average person listening to more music now than they were 40 years ago? If Western listening habits are stable, the inherent energy costs of the lifecycle of music production and distribution should be significantly reduced – moving bits (data) is always less energy-intensive than moving atoms (physical product). In the seventies and eighties, cassette tapes (the file sharer’s equivalent medium of choice), records and players all had embodied carbon and moving, rotating parts.

TRANSMISSION COSTS

While there may be some indicative energy costs of transmission, research on the breakdown of such costs, their significance and approaches to improvement remain sparse. Aligned to this, is there an end-point whereby data quality is compromised by the drive for energy efficiency?

SOCIAL BEHAVIOURS

While our instinct may be to impose top down legislative solutions, we have to be careful. There are unknown and unintended consequences as a result of imposing conditions within areas we would wish to change. For example, will data centres remain within countries that legislate against carbon consumption? If not, this could place the UK and its ICT industry potentially at risk from further globalisation, with the UK losing significant market share.

KEY CONSIDERATIONS FOR FUTURE RESEARCH:

Iain has a PhD in sustainability decision-support and is a council member for the Institute of Environmental Management and Assessment. He currently leads on sustainability and environmental policy matters for a local authority in Scotland, is an active public speaker on global sustainability issues and is an environmental auditor for A Greener Festival.

This paper suggests significant sustainability risks for both businesses and, with predicted two to four per cent global CO2 contributions, humanity’s viability on the planet. While alarming, its purpose is to engage the digital music industry and raise the profile of this issue, hence its scope narrowly focusing on the scale-up of current ‘business as usual’ digital media service provision.

However, the issue does not sit in isolation but is a part of the wider debate concerning ICT (data centre efficiencies and energy consumption), truly transformative business development (new distribution models and the challenges presented to content industries), corporate responsibility and consumer behaviour.

With ever-increasing government legislation acting to reduce carbon emissions, Data centre energy costs are expected to rise significantly more than inflation over the next few years which in turn are likely to affect margins throughout the supply chain. The ICT sector has been well aware of this for many years, with considerable research applied in order to reduce energy and associated carbon costs.

It is useful to note that with applied research, only half of the predicted data centre power usage actually materialised in the last decade. Nevertheless, there remains a question as to how far technology companies can achieve such efficiencies. Aligned to this is Jevon’s famous paradox1: that in certain areas, increased efficiencies are precursors to increased usage.

So-called rebound effects are inherent in many different areas of sustainability research and the industry should be mindful that the fears that this paper rightly raises could be realistic.

There are many questions that remain unanswered and this is part of the reason for necessarily widening the debate within this report.

IAIN HOSSACK – SUSTAINABILITY CONSULTANT

1 The proposition that technological progress that increases the efficiency with which a resource is used tends to increase (rather than decrease) the rate of consumption of that resource.

8 9

This report references a much larger 2010 Norwegian Research Council Report conducted by Bach Technology AS, University of Bergen and the Fraunhofer Institute for Digital Media Technologies as well as data from the original project proposal, which was submitted in 2009 (see Appendix 1).

The quantity of energy consumed in manufacturing memory cards, mobile handsets and other devices, the sources of that energy and the degree to which recycled components were used, are necessarily beyond the remit of this report, which unashamedly focuses on the energy consumption of data traffic.

While data traffic from streaming video continues to grow at an alarming rate and now accounts for the largest share of global data traffic, this paper concentrates on possible energy cost improvements to audio streaming and downloading services only.

A comprehensive analysis of the energy costs and solutions relating to streaming video is of the utmost importance, but beyond the scope of this work.

References to ‘file sharing’ should be taken to mean both licensed and unlicensed, unless otherwise specified. Identifying and distinguishing between the two is beyond the remit of this report; the wider issue of concern is the total volume of data traffic, however caused.

In terms of energy calculations regarding household/world/combined consumption, the report refers to the most recent data available at the time of writing – World Energy Report 2010. Figures for 2011 won’t be published until the very end of 2012.

ASSUMPTIONS UNDERSTANDING STORAGE CAPACITY

CAPACITY EQUIVALENT EQUATES TO

1 Megabyte (MB) 1,000 kB (kilobyte) 873 A4 pages of plain text

1 Gigabyte (GB) 1,000 MB 7 minutes of HD TV video 20 yards of books 256 MP3 audio files*

1 Terabyte (TB) 1,000 GB 262,144 MP3 audio files* 1,613 CDs**

1 Petabyte (PB) 1,000 TB 13 years of HD video 20 million 4-drawer filing cabinets filled with text 268,435,456 MP3 audio files* 1,651,910 CDs**

1 Exabyte (EB) 1,000 PB 35,000 years of HD video 274,877,906,944 MP3 audio files* 1,691,556,350 CDs**

1 Zettabyte (ZB) 1,000 EB 360 million years of HD video 281,474,976,710,656 MP3 audio files* 1,732,153,702,834 CDs**

1 Yottabyte (YB) 1,000 ZB 288,230,376,151,711,744 MP3 audio files* 1,773,725,391,702,841 CDs**

Figure 1

*4MB average file size

** 650 MB CDs

SOURCE: www.computerhope.com/issues/chspace.htm

10 11

The following data gives a snapshot of the forecast breakdown of total data traffic as defined by technology company Cisco’s industry-standard categorisations, using 2012 (fig 2: 2012) as a current point of reference and a 2015 projection (fig 2: 2015).

As shown below, the combined activities of file sharing, online video and web activity account for an overwhelming majority share of all data traffic combined.

Figure 2 shows the majoriy of data traffic is attributed to online video. While there is a modest downward trend projected in file sharing and web traffic, projected growth in online video (an increase of approximately 20 per cent by 2015) more than compensates for this decline.

1. DATA TRAFFIC AND ENERGY CONSUMPTION

1.1 SCALE AND COMPOSITION

File sharing

Internet Video

Web including downloads & streams

Other (video calling, online gaming & VoIP)

2012

2015

29.73%

49.62%

16.94%

3.71% 23.70%

57.75%

14.76%

3.79%

BREAKDOWN OF DATA TRAFFIC

Figure 2

The first compared the energy consumption of digital content delivery over the Internet with that of transportation of physical content. The results illustrated that online delivery can perform better in terms of energy consumption and carbon footprints, despite advances in greener transportation.

The other two analysed the effect of optimal implementation of streaming algorithms that can result in improved energy consumption. The second focused on communication protocols and their implementation, and the third on decoding of compressed digital content.

These studies have enabled a comparison to be made between the various transfer protocols with respect to energy consumption, bearing in mind that new protocols will appear which may have a positive impact on the energy consumption from data transfer.

The following reports have been used as references for the purpose of developing a methodology for calculating the energy consumption for transporting media content to a client and decoding it in the client device:

1. SHIPPING TO STREAMING: IS THIS SHIFT GREEN?2 – addressing the environmental and energy-related impacts of movie content delivery both via DVD and the Internet.

2. APPLICATION-SPECIFIC NETWORK MANAGEMENT FOR ENERGY-AWARE STREAMING OF POPULAR MEDIA FORMATS3 – addressing energy consumption and potential reduction of streaming via wireless networks.

3. A LOW POWER MPEG I/II LAYER 3 AUDIO DECODER4 – addressing energy consumption and possible reduction of the MPEG decoders used for receiving and converting a stream/download of audio/video.

2 Anand Seetharam, Manikandan Somasundaram, Don Towsley, Jim Kurose, Prashant Shenoy, (Department of Computer Science, University of Massachusetts, Amherst MA 01003 USA), 2010

3 Surendar Chandra (University of Georgia) and Amin Vahdat (Duke University).

4 Hyundai Electronics, Inc.

1.2 ANALYSIS

12 13

*10 Watt low-energy light bulb burning in hours (h) or minutes (m)

ENERGY CONSUMPTION COMPARISONS OF ONE SONG (3MIN/180SEC) IN PHYSICAL AND DIGITAL FORMATS, AND THEIR ‘LIGHT BULB* HOURS AND MINUTES’ EQUIVALENT

1 x download/stream of one song (192 kbps x180 sec./8 bits = 4.32Mb)Energy (Wh) 0.16

12 x download/stream of one song (12 x 192 kbps x180 sec. /8 bits = 51.84Mb)Energy (Wh) 1.97

1 x download/stream of one uncompressed WAV file (1,411 kbps x180 sec./8 bits = 31.75Mb)Energy (Wh) 1.21

12 x download/stream of one uncompressed WAV file (12 x 1,411 kbps x180 sec./8 bits = 381.02Mb)Energy (Wh) 14.55

Manufacturing and shipping of 12 songs (one CD album)Energy (Wh) 387.77

PHYSICAL

DIGITAL

Figure 3 SOURCE: Data compiled from Shipping To Streaming: Is This Shift Green?, Application-Specific Network Management For Energy-Aware Streaming Of Popular Media Formats and A Low Power MPEG I/II Layer 3 Audio Decoder

In 2010, Julie’s Bicycle, an organisation that addresses the environmental sustainability of the creative industries, produced the report How Green Is My Promo?5, looking into the distribution of promotional copies of music. It concluded that switching to the digital delivery of promo copies across the independent record sector would save 1,525 tonnes of CO2 annually – a reduction of 86 per cent.

So how does this relate to physical vs digital consumption of music itself?

Pooling the three research reports, it is possible to make energy consumption estimates for distribution of physical vs digital audio formats via various transfer protocols (fig 3).

1.3 DIGITAL VS PHYSICAL

5 How Green Is My Promo?, 2010

Repeated streaming of individual tracks may not necessarily be a desirable long-term solution with respect to energy consumption for the life cycle of a sound recording.

14 15

At first glance the energy consumption data in figure 3 appears to support the notion that digital streams and downloads are more energy efficient formats than their physical counterparts. While this is almost certainly the case with per-track downloads, taking these streaming figures at face value misses potentially bigger issues – repeated use across the lifetime of a consumer and increased bandwidth.

Streaming can result in favourite tracks being accessed potentially many hundreds of times throughout the lifetime of a listener.

While there’s a huge difference between energy consumed by one streamed MP3 (0.16 Wh) and that of an uncompressed track (1.21 Wh), advances in network infrastructure speeds in excess of 100 mbps (wireless) mean that audio files would no longer need to be compressed in future.

Thus streaming or downloading 12 tracks, without compression, just 27 times by one user would, in energy terms, equate to the production and shipping of one physical 12-track CD album. By extension, exceeding 27 streams or downloads per track would result in a greater energy and network burden than its physical counterpart.

It would therefore appear that repeated streaming of individual tracks may not necessarily be a desirable long-term solution with respect to energy consumption for the life cycle of a sound recording.

This is particularly true in the case of subscription models in which there are no financial incentives for the consumer to limit the number of streams.

Furthermore, increased bandwidth combined with better audio quality will lead to ever higher levels of energy consumption, due to the increased number of bits transported for the same listening session.

It is also important to emphasise the fact that these energy consumption figures (figure 3, P13) show an energy consumption rate 10 times lower than those found in a previous study conducted some years earlier6. In that respect, these energy consumption figures (above) are conservative.

Given that streaming formats increasingly signpost richer, deeper seams of related digital content (typically video streams and other rich media), it is likely that they will lead to an increasingly heavy energy burden.

If including streaming and cloud solutions for remote storage and repeated streams for every playback, (including streaming video from services like MUZU, VEVO or YouTube), there can be little doubt that the move to digital distribution of video and audio consumes vast amounts of energy and bandwidth. This is likely to be accelerated by the additional exponential growth of rich media associated with the original single track or album.

There can be little doubt that the move to digital distribution of video and audio consumes vast amounts of energy and bandwidth.

6 Another methodology to estimate the electricity intensity of data downloaded over the Internet, first presented in Koomey et al. (2004) and further developed by Taylor and Koomey in 2008 for Microsoft and Intel, estimated the electricity intensity of information transfers in kWh per gigabyte for 2000 and 2006, and from there updated the 2006 estimate to 2008. This study assumed an average electricity intensity of Internet data flows of about 7 kWh per gigabyte transferred for 2008 and 3.5 kWh for 2010.

Aside from many label and artist concerns regarding piracy, there may well be a heavier cost relating to the traffic over unlicensed P2P networks – surprisingly high energy consumption.

A German Ipoque study7 from 2007 concluded that 70 per cent of the data traffic in Europe was from unlicensed file sharing traffic.

Recent Cisco forecasts8 take this further, showing continued growth of file sharing traffic within both terrestrial and mobile networks, albeit reducing to 40 per cent of all data traffic in 2010 and to 24 per cent in 2015, mainly because of the rapid growth of online video9.

Data traffic consumption from P2P networks rose from around 10 per cent in 2000 to over 70 per cent of all Internet traffic in Europe 200710. If unlicensed file sharing represents 70 per cent of this 70 per cent, it may consume 1 per cent of the world’s energy, equivalent to 300 billion kWh or approximately 32 million kWh, every day (24/7) for a year11.

Put another way, unlicensed file sharing could consume the equivalent of up to four times12 the annual combined electricity consumption of all UK households13.

When the IPRED14 anti-piracy law was introduced in Sweden in April 2009, allowing the authorities to collect the IP addresses of unlawful file sharers, web traffic almost immediately decreased by 33 per cent – from 120 GBps to 80 GBps15. This indicates that this quick reduction was probably due to a drop in file sharing.

The reduction itself was equivalent to 158 petabytes per year (a digital version of all publicly available books in the world in all the world’s languages corresponds to 0.4 petabytes), resulting in a reduction in energy consumption of 6,000 MWh per year in Sweden (based on ADSL network energy consumption)16 – the equivalent to the annual combined electricity consumption of 2,030 UK households.

The fact that a good proportion of these ex-file-sharers most likely moved to Spotify would suggest a more permanent reduction in energy consumption from their behaviour. This is mainly because Spotify introduced a cache solution (in combination with a torrent protocol) to avoid the repeat streaming of the same songs, and also it and its competitors have far more sophisticated search/filtering solutions due to better quality metadata than file sharing networks, enabling better targeted searches and more accurate results. Thus users are able to precisely locate a piece of music and avoid inadvertently downloading multiple files of the same recording.

2.1 UNLICENSED FILE SHARING

2. DIGITAL CONTENT

7 Ipoque Studies 8 Cisco Visual Networking Index, 20119 Beginning with the launch of YouTube in 2005, online video achieved

significant scale by 2007, aided by the launch of the iPhone, triggering streaming on mass-market smartphones. 2009 marked the point at which video represented a bigger share of data traffic than file sharing. At the same time, both the bandwidth of mobile networks and the storage capacity of memory cards for smartphones were dramatically increasing.

10 Ipoque Studies11 World Energy Outlook 2010 and US Energy Outlook 201012 ONS 2012 – Key findings reveal approximately 26.3 million UK

households in 201113 Based on a calculation from Carbon Footprint14 IPRED - The Intellectual Property Rights Enforcement Directive - A long-

running proposal by the European Union to increase penalties and ultimately criminalise ‘commercial’ infractions of intellectual property law within the EU.

15 GPRS measurements are in gigabits per second16 World Energy Outlook, 2010

16 17

Licensed downloading (whereby the end-user legally downloads a song or album and keeps it in perpetuity on their computer/portable device) started to impact the recorded music market around a decade ago and is typified by services such as iTunes, Amazon MP3 and eMusic. In essence, the user has a lifetime licence to listen to downloaded songs as many times as they wish, although some services place restrictions on what devices downloads can be played on and how many times they can be copied or burned onto CD.

On-demand streaming is a relatively new form of delivery for mainstream consumers, primarily due to the bandwidth required to ensure tracks stream at a consistently high quality without glitches or buffering problems.

Over recent years, driven by the advent of ‘smart’ devices and highspeed broadband, a variety of audio streaming services have gained in popularity. Several of these offer an offline caching solution to avoid repeated streaming of the most popular tracks. This sees files downloaded and held on registered devices for unlimited playback for the duration of a subscription and without need for a web or mobile connection after the initial download. This is helping to reduce data traffic.

2.2 DOWNLOADING AND STREAMING

Whereas cache-based audio streaming services can use a device’s RAM to store relatively small file sizes (typically 10Mb), as yet, video services can’t meaningfully exploit cache-based solutions due to file sizes often being too large – as an example, YouTube cap uploaded files at 2GB.

One option might be to introduce limited caching from the ISP delivering streamed video to the end users, but, because of the distance from the ISP’s servers to the end user, this would not significantly reduce the volume of data traffic.

Further, new models and more sophisticated recommendation tools encouraging increased exploration of catalogue tracks, which may only be played once yet be cached regardless, place further restrictions on the cache’s storage capacity.End-user music consumption behaviour should be further researched in order to optimise cache-based solutions.

3. TECHNOLOGY

3.1 WIRELESS VS TERRESTRIAL

17 There are two standards for Bluetooth: minimum and maximum.

Given the rapid shift from terrestrial (ADSL or similar protocol using the current copper lines in the ground) to wireless networks, it is important to compare these two alternatives from an energy perspective.

Figure 4 comprises data pooled from the three reports mentioned in section 2. This shows that WLAN

(Wireless Local Area Network) and Bluetooth are the most energy-efficient wireless alternatives. However, WLAN often still requires a fixed-line ADSL connection to the WLAN router resulting in bandwidth limitations and increased energy consumption. Bluetooth will only work with devices in close proximity, e.g. in the same room, and therefore is only suited to close proximity ‘digital handshake’ exchange of individual small files.

MICROSOFT MEDIA WLAN (128KBPS) for two clients (excl. decoding). Mostly terrestrial - only the last metres from the router are wireless)

GPRS (128kbps x180 seconds or 2,880 MB excl. decoding)

0.054 kJ (0.015 Wh)

0.135 kJ (0.038 Wh)

0.003 kJ (0.001 Wh)

0.232 kJ (0.064 Wh)

0.023 kJ (0.006 Wh)

DECODING (180 seconds MPEG/II Layer 3, all bit rates) (in the device)

BLUETOOTH MAXIMUM

BLUETOOTH MINIMUM

Figure 4

18 19

In general, the various wireless solutions and their respective energy consumptions are directly related to the distance between transmitters.

WiMAX transmitters (Worldwide Interoperability for Microwave Access – the newest and most energy-efficient data transfer protocol available today), typically achieve a reach of 50 km, while a GPRS/3G transmitter (General Packet Radio Service) has a reach of just 20 km, resulting in more GPRS/3G transmitters being required over a given distance than WiMAX, making GPRS/3G transmission less energy efficient18.

Figure 5 has incorporated this factor as well as breaking down the energy needed for decoding and streaming a three minute media file (audio or video) at bitrates of 128 kbps and 192 kbps19. (It’s worth clarifying that it doesn’t matter whether the file is audio or video, if the bitrate is the same, then the three minute files will contain the same amount of bits, i.e. the same amount of data. And yes, at the bitrates suggested we would be talking about rather grainy video!).

WLAN 128 kbps 180 seconds (wireless + terrestrial backbone)

GPRS 128 kbps 180 seconds (wireless)

WiMAX (wireless)

BLUETOOTH MAXIMUM

BLUETOOTH MINIMUM

ADSL 192 kbps 180 seconds (terrestrial)

Figure 5

18 Interview with Patrick Aichroth, Fraunhofer Institute For Digital Media Technology, Dept for P2P & Security, 10 April 2012

19 A Low Power MPEG I/II Layer 3 Audio Decoder, SM Team, System IC, Hyundai Electronics Inc.

0.655 kJ (0.182 Wh)

0.138 kJ (0.038 Wh)

0.597 kJ (0.166 Wh)

0.037 kJ (0.010 Wh)

0.235 kJ (0.065 Wh)

0.026 kJ (0.007 Wh)

If anticipated data traffic continues to increase following linear and not an exponential growth, data traffic is likely to reach the threshold of 1 yottabyte by 2027.

ENERGY CONSUMPTION REQUIRED TO DECODE AND STREAM A 3 MINUTE MEDIA FILE

20 21

Given that the number of people with access to the Internet is shortly expected to reach 2 billion globally, if we divided them equally between wired (ADSL 192 kps) and wireless (GPRS 128 kps) networks and assume that each person downloads or streams music for 2 hours per day (without caching), this would

equate to the annual electricity consumption of more than 2 million UK households.

Pooling data from the Cisco report20 and the three reports referred to in section 2, figure 6 details the various data volumes (in bytes) and how much energy they consume over various networks.

VOLUME OF DATA DATA TRANSFER ENERGY CONSUMPTION OVER DATA NETWORKS

CAPACITY EQUIVALENT ADSL WLAN GPRS WIMAX

1 megabyte (MB)

1,000 kB (kilobyte)

0.0384 Wh 0.0631 Wh 0.0133 Wh 0.0036 Wh

1 gigabyte (GB)

1,000 MB 38.39 Wh 63.11 Wh 13.34 Wh 3.60 Wh

1 terrabyte (TB)

1,000 GB 38.39 KWh 63.11 KWh 13.34 KWh 3.60 KWh

1 petabyte (PB)

1,000 TB 38.39 MWh 63.11 MWh 13.34 MWh 3.60 MWh

1 exabyte (EB)

1,000 PB 38.39 GWh 63.11 GWh 13.34 GWh 3.60 GWh

1 zettabyte (ZB)

1,000 EB 38.39 TWh 63.11 TWh 13.34 TWh 3.60 TWh

1 yottabyte (YB)

1,000 ZB 38,392 TWh 63,110 TWh 13,337 TWh 3,600 TWh

20 Cisco Visual Networking Index 2011

21 WiMAX Forum Forecast 201222 World Energy Outlook 201023 Cisco Visual Networking Index 2011

Cisco 2015 projections of data traffic cite a threshold of 1 zettabyte (Appendix 2). This would be equivalent to 38.39 TWh per year on ADSL, or an energy consumption equivalent to 0.2 per cent of the world’s total electricity consumption in 2010. If anticipated data traffic continues to increase following linear and not an exponential growth, data traffic is likely to reach the threshold of 1 yottabyte by 2027.

Even with all traffic moving over to WiMAX, this traffic will nevertheless consume the energy equivalent of 21 per cent of the world’s total electricity consumption in 2010.

The reality is worse, however. With WiMAX being a leapfrog technology, while it is likely to become more widespread in less technologically developed markets, it is less likely to meaningfully penetrate technologically advanced regions such as Western Europe and the US for the foreseeable future21.

To put this into context with respect to traffic volume, 1 petabyte of data is equivalent to three AAC downloads per second, every hour of every day for one year; while 1 zettabyte is equivalent to 3 million AAC downloads per second, every hour of every day for one year.

While many policymakers around the globe are focused on the sustainability of energy production and usage, it seems doubtful that the energy cost of the rapid growth of data traffic will be compensated for by new technological achievements. Therefore, these forecasts should be taken very seriously with respect to developing less intensive online data transfer services.

To further illustrate the scale of data traffic and its energy drain, 2011 YouTube statistics indicate some 4 billion video streams per day (see Appendix 2). Assuming a 1GB file size per video (half of YouTube’s 2GB cap), this represents daily data traffic consumption of approximately 8 exabytes – annually equivalent to 0.1 per cent of the world’s electricity consumption in 201022.

With Google recently reporting a 25 per cent growth rate in just eight months (Appendix 2), energy consumption from YouTube video delivery will, by 2013, annually equate to one per cent of the world’s electricity consumption in 2010 (this assumes no caching, however, and that the most efficient transfer protocol currently available is being used, which is highly unlikely).

Figure 6

22 23

The mobile entertainment business is greener than the conventional entertainment business by miles because there is no packaging and, for the most part, everything is digitally downloaded. From that standpoint, it is energy efficient. It is a much cleaner and much more direct way of getting to the consumer. In terms of the upgrade cycle on handsets, in Japan it’s every 10-12 months compared to every 18-24 months in the West but from a mobile entertainment perspective, that’s not really a factor. It’s similar to the tablet and laptop industries.

Mobile handset makers and the GSM Association, the global trade body, have a very comprehensive environmental and recycling policy and this is certainly deployed by most, if not all, of the handset makers. Nokia for one has a very strict and determined environmental policy, including the effective gathering in and disposal of used devices. All of the manufacturers, as far as I know, have a considerable eye on sustainable manufacture and environmental concerns. Insofar as the use of sustainable energy resources, every manufacturer is looking at maximising battery performance on new devices and finding new ways to power up – e.g. in Africa, solar panels have been used extensively in certain countries to power up mobile phone devices without having to resort to using electric power from the grid.

Compression technology is a constant question for everybody here as technologies that can squeeze additional capabilities out of a network will enable data networks and telco infrastructures to process more and subsequently offer more content-heavy services. The big development is NFC (near field communications) technologies, which allow high-speed 4G data transmission in dense urban environments.

RALPH SIMON, CHAIRMAN EMERITUS, MOBILE ENTERTAINMENT FORUM ON MOBILE

How green is the mobile entertainment sector?

Can more be done to recycle redundant handsets and are device manufacturers and mobile operators themselves looking to increase their use of sustainable energy sources?

As file sizes get more compressed and mobile internet speeds get faster, will this merely make smartphone users consume more?

The mobile entertainment sector worldwide now is a $25 billion business. It’s still way behind the videogame business but it is on a par with the traditional record business. In terms of mobile data, voice is the biggest revenue stream and second is text messages. The largest configuration of entertainment properties is caller ringback tones. That is followed by news, weather and sport. Then there are mobile games. In terms of streaming music, consider that there are 6 billion mobile phone users and Spotify currently has 20 million users per month; even while many of those phones aren’t smartphones and so cannot run Spotify’s app, that gives you some idea of the potential.[Spotify currently has 0.167 per cent of the potential mobile streaming market.]

Africa has leap-frogged fixed-line technologies and the projection is that there will be 1 billion people in Africa with mobile phones in the next five years. India will hit the 1 billion subscriber mark and China has already exceeded that. The telcos are thinking about how they can focus on these markets because of the huge pick-up.

It’s very small. Only about 5 per cent of phones in the BRICA nations are smartphones. The rest are feature phones and entry-level devices. There is a move to low cost smartphones being introduced in places like Africa that are opening up mobile internet to a whole new constituency for the first time. There are tablet devices that will retail for $38 in India to be used in education establishments. The biggest development in Africa is the connection of the high-speed broadband under-sea submarine cables down the east and west coasts of the continent. In the next 18 months when it’s connected, it will raise the broadband capability of Africa to new levels, which will in turn enable mobile Internet.

It seems in Western markets that is the case. The battle for subscription models is between the likes of Spotify, Deezer, Simfy, Slacker and Rhapsody. Emerging markets are going to be dominated by IVR (interactive voice response) rather than streaming for the next 12-18 months. IVR is where you dial a number and use a menu on your phone. So, say you wanted to listen to the top 10, you choose option ‘5’.

Ralph SimonSeptember 2012

How much of mobile entertainment data does music command?

How will booming mobile usage affect the BRICA (Brazil, Russia, India, China, Africa) nations?

Are smartphones taking off in developing markets?

Will users and services move towards streaming rather than downloading?

24 25

Internet World Stats reports that the highest number of Internet users are in Asia with 922 million users, while North America has the highest penetration with 78.3 per cent27. This will probably change quite rapidly, with new networks and the rapidly increasing number of smartphones.

Including machine-to-machine (M2M) devices Cisco further predicts28 that there will be over 7.1 billion mobile-connected devices, in 2015 – approximately equal to the world’s population in 2015 (7.2 billion).

Thus we are potentially talking about a future scenario where there are just as many mobile connected

devices as there are global inhabitants and it would be reasonably safe to assume that most of the world’s population (or at least the majority of households) will therefore potentially have access to the Internet, with the majority achieving this via smartphones.

This sets out plainly that the expected data traffic growth and its corresponding energy cost will have serious consequences, unless solutions are put forward for significantly more energy efficient data transfer protocols.

3.3 GLOBAL MOBILE INTERNET PENETRATION

3.2 MOBILE COMMUNICATION

It has been projected that there will be 3 billion mobile internet users by 201523. Infonetics Research reported in 2011 that the number of cellular mobile broadband subscribers jumped almost 60 per cent in 2010 to 558 million worldwide24 and should exceed 2 billion by 2015. Meanwhile, a Greenpeace report25 predicts that target may be achieved as early as the end of 2012.

However accurate these statistics are, one thing is certain – mobile handset ownership and smartphone access to the Internet is booming at an unprecedented rate.

A new data transfer protocol standard – MGMN (Multiple Gateway in Mobile Network) – supports download speeds of 100 mbps which allow the download of four AAC files per second and streaming YouTube HD onto a smartphone or tablet. If this protocol becomes a new standard (it’s currently too early to say), sources in the European Commission have been looking at exponential projections of data traffic growth26. What is certain is that new protocols will support uncompressed, larger files with a corresponding drain on infrastructure, traffic and energy.

24 Infonetics 201125 Make IT Green: Cloud Computing And Its Contribution To Climate

Change 201026 Dr. João Schwarz da Silva, Director of the Converged Networks and

Services, EU Commission, Keynote speech, VERDIKT Conference, Bergen 2008, predicted that data traffic could be doubled every hour after 2020

27 Internet World Stats28 Cisco Visual Networking Index 2011

The last couple of years have seen healthy growth in streaming services, as well as the development of locker and scan and match services whereby users can upload their files (both legally and illegally acquired) and pay a subscription, from which the rights holders receive a share of revenue. The scan and match services use digital fingerprints from a verified fingerprint database for recognising uploaded files and for the tracking and reporting of usage29.

Competing with unlicensed services through providing better filtering (more accurate search results) and higher quality, secure files might be a draw for the consumer, and will most likely reduce the massive overload of data traffic from more primitive unlicensed file sharing services, whereby the user simply enters the name of an artist and downloads all the (potentially) hundreds of files located.

The section below considers a couple of hypothetical music distribution scenarios in order to illustrate some of the ways in which future models might become more energy efficient.

ALL YOU CAN EAT AND SHARE

The first of these scenarios is an ‘All You Can Eat and Share’ model at an affordable subscription fee. This might comprise a tethered download service, or combine download and streaming within the same service.

Theoretically, as mentioned earlier by Ralph Simon, a potential 6 billion mobile subscribers each paying a moderate and affordable subscription fee has the potential to generate more revenue for the music industry than it could ever have generated through physical sales.

The downside to this is that in being affordable and unlimited, the volume of data traffic would be even greater than a streaming music worst-case scenario, resulting in uncontrolled energy consumption whereby energy providers stand to profit more from music than rights holders themselves.

This supports the argument that a progressive download or cached service (see section 1.2) or a ‘close-to-consumer’ cloud solution might be the most environmentally-friendly option, assuming licensing issues could be resolved. These technologies consume less energy and in such a scenario, a close-to-consumer cloud could be located in your city, your neighbourhood, or even in your house.

Energy efficiencies could also derive from the ‘share’ part of the model, whereby, like torrenting, the service locates files closest to the end user. The shorter the journey the more energy-efficient the transfer. In the case of Bluetooth transfer of files between friends, the efficiencies would be considerable. Once again, the ‘share’ part of the model is being suggested from the point of view of energy-efficiency alone, with no consideration of licensing sensitivities.

4. SUSTAINABILITY

4.1 SUSTAINABILITY FOR THE RIGHTS HOLDERS

29 In 2011, Apple launched its own iTunes Match subscription service which, as part of its wider iCloud offering, matches a user’s collection in the Cloud (including, in theory, tracks a user bought from iTunes as well as ripped or downloaded from legal or illegal sources) for access via any registered iOS devices (iPhone, iPad, etc.).

26 27

PRE-LOADING MUSIC

The second scenario is based upon the energy-efficiencies of physical storage, with its rapidly falling costs, ever-increasing capacity.

The pre-loading of millions of tracks on a chip that users pay to unlock content from (rather than stream or download files) would reduce data traffic and associated energy consumption significantly as the only data costs would be those relating to metadata required to unlock the music, rather than the files themselves.

The pre-loading of memory cards already takes place among pirates in countries like India and China, with the entire recorded output of artists like the Beatles available on high-capacity cards for a few dollars. Should memory card capacity increase as forecast (section 4.3), the recorded music sector potentially faces a bleak future, in which it would at least theoretically be possible for a pirate to copy and distribute all known recorded music in its entirety, in any language.

The industry could take a lead on the development of this to combat piracy losses, offering low cost, high quality memory cards for use in mobiles and other devices, revenues from which might be greater than from any other physical format.

Clearly the notion that the world’s entire recorded output could be stored on a chip and ‘unlocked’ by a subscriber marks a step-change as a distribution model, posing significant security and licensing issues.

Ralph Simon again:

Pre-loading is still very much in its infancy. It’s not really a factor of consequence at the moment – but it might be in 12 months’ time. There have been previous attempts. Nokia tried it with ‘Comes With Music’, which was not successful. It remains to be seen whether Spotify can be expanded at a fast rate. Preloading might become more important in 12–18 months, but right now it’s still a bit premature.

Below are three possible ways of helping develop more energy efficient digital distribution platforms in the future:

1.REDUCE THE DATA PROCESSING POWER FOR TRANSMISSION AND MONITORING Move away from traditional terrestrial to mobile or fibre/Wi-Fi infrastructures that consume less energy.

2.REDUCE THE DATA TRANSMISSION RATE OF SHARED CONTENTMove towards mobile distribution, ape the ‘local’ BitTorrent protocol to move towards regional and local clouds as hubs for middle storage between a central service and the end-user. Pre-loading could become the ‘near’ cloud.

3.REDUCE FILE SHARING AND THE AVAILABILITY OF IDENTICAL MULTIPLE FILES via platforms which return more accurate search results and contain fewer duplicate files residing in cyberspace, resulting in reduced bandwidth usage.

4.2 ENERGY EFFICIENT DISTRIBUTION

Should memory card capacity increase as forecast, the recorded music sector potentially faces a bleak future, in which it would at least theoretically be possible for a pirate to copy and distribute all known recorded music in its entirety, in any language.

28 29

Moore’s Law30 proposes that every year the storage capacity of computer memory chips doubles while their costs halve, ultimately leading to a point where storage solutions can be arrived at a near-zero price.

To illustrate this, in 1960 the estimated computer data storage cost was $8 per byte. Based on that, a standard 32GB mobile phone memory card would now cost $256 billion.

Thus we are likely to see a similarly dramatic decrease in storage costs in the coming years, whereby a 1 petabyte hard drive could soon cost $100 and be able to store all the songs in the world in the AAC or FLAC formats – or even 30 million songs as WAV files. To put that in context, to listen to all the songs held on a 1 PB drive would require listening to music solidly for 12 hours a day for 342 years.

In 2010, SanDisk released the SanDisk Ultra SDXC containing up to 64GB of storage. At the time, this was hailed as the highest capacity SD memory card ever. Announced late 2010/early 2011, was a new SDXC 2TB memory card that will enable storage of more than 600,000 AAC files on a mobile phone (equal to eight years of constant listening over 12 hours per day31). Though these have yet to materialise, it is a matter of ‘when’, not ‘if’.

The new CompactFlash 5.0 standard promises a theoretical 144 petabytes of storage capacity for newer CompactFlash cards, equating to 28.8 billion AAC files or 4.3 billion WAV files.

Figure 7 gives an overview of forthcoming memory card storage capacities, indicating that there will be no storage capacity issues preventing a platform from delivering a combination of pre-load or progressive download content, as discussed in section 4.1.

4.3 A SOLUTION FOR THE FUTURE?

30 Moore’s Law is the observation that over the history of computing hardware, the number of transistors on integrated circuits doubles approximately every two years.

31 Are SDXC Cards the Future of Music Storage? 2012

2,000 1,538400,000117,64760,606

GbytesHD moviesAAC songsFlac songsWav songs

=

1,000,000 769,231200,000,000(14,285,714)58,823,529(4,201,681)30,303,030(2,164,502)

GbytesHD moviesAAC songsalbumsFlac songsalbumsWav songsalbums

144,000,000 110,769,23128,800,000,000(2,057,142,857)8,470,588,235(605,042,017)4,363,636,364(311,688,312)

GbytesHD moviesAAC songsalbumsFlac songsalbumsWav songsalbums

100,000 76,92320,000,000(1,428,571)5,882,353(420,168)3,030,303(216,450)

GbytesHD moviesAAC songsalbumsFlac songsalbumsWav songsalbums

MEMORY CARD STORAGE CAPACITIES

2TB SDXC MEMORY CARD

100TB SDXC MEMORY CARD

1 PETABYTE HARD DRIVE

144 PETABYTE (COMPACT FLASH MAX STORAGE)

Figure 7

30 31

5. CONCLUSIONS

15 years ago, the energy consumption and associated environmental costs of digital distribution were hardly a consideration. The more traffic grows, the more important it is for environmental sustainability to be part of the equation.

Ultimately, this report sheds light on only part of a very complex picture.

To complete it, a huge research undertaking is required that, among other things, quantifies the energy consumption of the manufacturing process for mobiles, devices, transmitters, peripherals, storage media and network infrastructure

There’s also a job of work to be done in quantifying the energy consumption that device upgrading, usage patterns and charging habits incur.

Corporate responsibility with regards the sourcing of energy and the provenance of materials used in device manufacture, hardware and software compatibility and product shelf-life are, likewise, key.

So too is consumer responsibility, to include charging habits, upgrading and usage patterns. And should the consumer be made aware of the energy cost of digital file access and transfer through their billing, thus enabling more informed choices to be made?

The recordings business once again finds itself at the vanguard of profound technological change and the adoption of more energy efficient protocols depends in part on its willingness to continually license and monetise content through new and evolving distribution channels. Only then will it be possible to more accurately evaluate the most sustainable forms of digital distribution.

There are four possible solutions with which to tackle the pressing energy issues as highlighted in this report, each of which deserves further study:

1) Is caching an efficient solution for reducing data traffic from streaming, and should this become obligatory for all streaming services?

2) Should the industry simulate the use of P2P protocols for cached streaming and downloads, whereby a user requesting content will be serviced from another user with the same content and from one who is geographically closest to the requesting user?

3) Should streamed data be restricted to new and dynamic info, while the core music file itself be preloaded or downloaded/cached once only?

4) Should data traffic be taxed in order to reduce data overload?32

Many of these points raise issues that need careful consideration, not least regarding which authority might enforce such restrictions globally. However, given the imminent arrival of very high density, low-cost data storage and a booming cloud infrastructure, the time has come for creative solutions to be explored, which will no doubt spur and inform others in the digital content sector.

32 Portugal has proposed a ‘Terrabyte Tax’, though it is not clear whether the purpose is an environmental one

High capacity storage cards aren’t set to hit the market for some time yet, for various reasons; yet-to-be-agreed standardisation between manufacturers; the unit cost of production (high) and price to the consumer (expensive); mobiles and portable devices don’t yet have the necessary processing power required for such cards.

However, looking back to 1989 when an Apple MacIntosh SE with a 20 MB hard drive cost £3,000 GBP, we can safely assume that they will appear when the market is ready.

It’s still a hypothetical scenario, but one possible sustainable future distribution model that reduces the volume of data traffic itself could lie with a combination of pre-loaded existing catalogue (static data) and one-time downloading only for new tracks (dynamic data).

From a security point of view, if these new memory cards do indeed develop in the way predicted, without static and dynamic data, cards could be pirated, making this a bigger potential problem than that of hacking data in the first place.

But, if the majority of commercially released songs to-date were pre-loaded onto a storage device without metadata, they would be useless for any other purpose other than random play. To unlock their value therefore, the

relevant metadata would need to be accessed online.

This gives rise to a further scenario, limiting streaming to dynamic data only:

• Sell metadata only to unlock pre-loaded catalogue (the 90 per cent above); in effect, reducing streaming to super-low bit transfer rates (with a commensurate saving in energy)

and

• Stream/download only new tracks, once, to each consumer (thus avoiding repeated streams of the same file from multiple sources).

The immediate benefits could be:

• Low-end phones and low bit rate networks could offer intelligent music services combining metadata streaming and pre-loaded content and subscriptions bundled with sales of mobiles and subscriptions to mobile services, especially in emerging markets.

• The online mobile market could grow rapidly to reach several billion end-users, which could include a minimum revenue guarantee for rights holders.

32 33

APPENDIX 1BACKGROUND

This paper is a condensed version of selected chapters of an internal sub-report of a three-year R&D project, Sustainable & Green Solutions For Online Media In Enhanced Networks for the Norwegian Council of Research (Project no. 201434/S10), conducted by Bach Technology AS, University of Bergen (Institute for Informatics, Dept. for Data Optimisation), and the Fraunhofer Institute for Digital Media Technologies (2010–2012).

The scope of that report did not include calculations of likely CO2 emissions of the music or digital media sectors, nor a deep analysis of the environmental aspects of digital distribution. Rather, its goal was to identify potential challenges and sustainable solutions with respect to uncontrolled data traffic growth coming from online music and media, where the starting point in 2009 – the stage at which the main R&D project was initiated – was addressing illegal file sharing.

To the best of the authors’ knowledge, while various papers have considered ICT energy consumption

overall (particularly issues around data centre and server consumption), notably less attention has been given to the energy cost of the data and network traffic itself, particularly in light of data traffic forecasts and capacity developments in memory cards.

An outline presentation of the R&D research report’s initial findings at a music industry roundtable discussion in Norway 2010 concluded that the future will indeed bring with it exponential growth in data traffic, large capacity, low-cost storage media and a transition from terrestrial (fixed-line and wireless) to mobile networks.

The natural progression from this report was to consider the energy impact of data traffic and how to meet the potential piracy threat posed by the imminent arrival of ultra-high capacity memory cards which will literally enable the storing of all known recorded music in the world on one device.

APPENDIX 2DATA CENTRES, DATA NETWORKS AND ENERGY CONSUMPTION

Data centres have long been the Internet’s energy hogs. A recent 2011 study, The 2011 Data Centre Industry Census by DatacenterDynamics35, estimates that the world’s data centres will consume 19 per cent more energy in the next 12 months than they have in the past year.

The report is a comparative study of data centre operators and end-users as well as the facilities which they manage. It estimates that data centres currently consume about 31GW, which represents 271 TWh per year or 1.6 per cent of the world’s electric energy consumption. Compared with the study from Gartner in 2009, this indicates that the main energy consuming factor within the ICT industry is data communication.

A Greenpeace report, Make IT Green…36 likewise highlighted the scale of the sector’s estimated energy consumption, and provided new analysis on the projected growth in energy consumption of both the Internet and cloud computing for the coming decade, particularly as driven by data centres.

Even though the predicted growth of 19–22 per cent more energy over the coming 12 months to be consumed by data centres is a pessimistic scenario, there are significant energy efficient improvements necessarily taking place with server technologies, simply because the electricity bills are ‘killing’ the business of the data centres.

According to the US Energy Information Administration, total world energy consumption in 2010 was estimated to be 150,000 TWh per year.33 Of this, electricity represented 17,000 TWh. The global ICT industry (Information, Communication and Technology) consumes approximately 3,000 TWh or 17 per cent

of the world’s electricity consumption and generates about 2 per cent of global carbon dioxide emissions. According to a recent Gartner study cited by IKT Norway in a report on green data centres in 200934, this contributes as much greenhouse gas to the atmosphere as the world’s airlines combined.

DATA CENTRES

33 1 TWh is 1 billion Kilowatts an hour; 150.000 TWh per year is equivalent to the annual electric consumption of 50 billion British households.

34 Sky Og Fjordane (Cloud and the Fjords), IKT Norway, 200935 DatacentreDynamics Research36 Make IT Green: Cloud Computing And Its Contribution To Climate

Change, 2010

34 35

Hewlett Packard (HP) recently announced the Moonshot project37 in which its next-generation data centres will reduce their physical size by 94 per cent and reduce their associated power consumption by 96 per cent. From this, HP expects 75 per cent savings in capital expenditure.

Within this strategy are recently-announced new-generation Power Optimised Datacentres (PODS) configured as 12-metre containers which can be regulated by water cooling or an advanced air-cooling solution, reducing the Power Usage Efficiency (PUE) from 2 to 1.0638. A data centre based on PODs and future generations of Moonshot servers is expected to give a 95 per cent

annual energy saving. In addition, many PODs are located close to hydro-electric power plants in order to achieve low power source CO2 emissions, combined with access to water for cooling the data centre. These technological advances could become important infrastructure components in a solution in which main clouds would copy content to regional clouds.

From a music industry perspective, the key consideration is whether this could be part of a wholesale move to regional clouds, evolving into local storage and even pre-loaded storage solutions (ref 4.2 and 4.3).

37 Houston, We Have A Solution38 PUE is a measure which describes the total power used for running

a data centre including infrastructure, power transport and cooling, divided by the power used for running the servers only. The optimal PUE is 1.

39 Cisco Visual Networking Index 2011

2010 2011 2012 2013 2014 2015

FILE SHARING* 39.60 33.68 29.73 26.89 25.22 23.70

INTERNET VIDEO* 37.20 45.22 49.62 53.33 55.65 57.75

WEB, EMAIL & DATA* 19.10 17.42 16.94 16.15 15.46 14.76

VIDEO CALLING 2.46 2.47 2.69 2.74 2.86 2.98

ONLINE GAMING 0.39 0.38 0.39 0.40 0.43 0.50

VOICE OVER IP (VOIP) 1.10 0.82 0.63 0.48 0.37 0.29

OTHER 0.00 0.01 0.00 0.01 0.02 0.02

100.00 100.00 100.00 100.00 100.00 100.00* Most relevant for the music industry

SOURCE: Cisco Visual Networking Index 2011

Cisco39 - the acknowledged provider of authoritative statistics throughout the ICT industry – define data traffic accordingly:

• FILE SHARING: includes peer-to-peer traffic from all recognised P2P systems such as BitTorrent and eDonkey, as well as traffic from web-based file sharing systems

• INTERNET VIDEO: includes short-form Internet video (for example, YouTube), long-form Internet

video (for example, Hulu), live Internet video, internet-video-to-TV (for example, Netflix through Roku), online video purchases and rentals, webcam viewing, and web-based video monitoring (excludes P2P video file downloads)

• WEB, EMAIL, AND DATA: includes web, email, instant messaging, and other data traffic such as streaming, download (excluding file sharing but including lockers)

DATA TRAFFIC

• VIDEO COMMUNICATIONS: includes Internet video-calling over instant messenger and soft-client video calling programs such as Skype

• GAMING: includes casual online gaming, networked console gaming, and multiplayer virtual-world gaming

• VOIP: includes traffic from retail VoIP services and PC-based VoIP, but excludes wholesale VoIP transport

Figure 8

The following table details projections of the per centage share of Cisco-forecasted global data traffic across these categories.

36 37

Cisco predicts that the global data traffic volume will reach a threshold of 1 zettabyte (ZB) by 2015. This equals 1 billion gigabytes or 1,000 billion megabytes. With an expected population of 8 billion inhabitants in the world this represents an annual data consumption of 125 megabytes per head of population, worldwide.

The same Cisco report gives an overview and prediction of what kind of services that are and will be dominating data traffic. From file sharing being the most dominant in 2009, there is significant growth towards streaming video becoming dominant with respect to data traffic, of which music video is a significant part.

With ever more sophisticated methods of packaging audio alongside high resolution photos, video and streamed links, (not to mention the audio quality itself becoming near lossless), the quantity of delivered bits per listening session of audio is almost certainly set to dramatically increase.

A 2012 announcement from Google that YouTube is now streaming 4 billion videos every day40 would appear to evidence this trend. This equates to a 25 per cent increase in the preceeding eight months,

largely attributable to YouTube expanding its reach beyond computers to smartphones and television devices. (YouTube is currently, and by some distance, the single biggest platform for online video, comprising 27 per cent of mobile Internet traffic in North America; Netflix accounts for just 2 per cent41).

Combining Cisco forecast figures for file sharing and internet video indicates a near-constant share of data traffic (77 per cent in 2010 through to 81 per cent in 2015), in turn creating the biggest challenge with respect to data traffic and energy consumption.

Taking total data traffic and Compound Annual Growth Rate (CAGR) the following table (figure 9, P38) illustrates projected growth within each category.

40 Reuters, 23rd January 201241 Techcrunch, 25th April 2012

With ever more sophisticated methods of packaging audio alongside high resolution photos, video and streamed links, (not to mention the audio quality itself becoming near lossless), the quantity of delivered bits per listening session of audio is almost certainly set to dramatically increase.

38 392010 2011 2012 2013 2014

50 000

40 000

30 000

20 000

10 000

12,528

17,867

24,477

32,973

43,772

58,214

VOLUME OF GLOBAL DATA TRAFFIC (PETABYTES) / YEAR (FORECAST)

2010 2011 2012 2013 2014 2015 CAGR

FILE SHARING

4,968 6,017 7,277 8,867 11,040 13,797 23%

STREAMING VIDEO

4,672 8,079 12,146 17,583 24,357 33,620 48%

WEB, EMAIL & DATA

2,393 3,113 4,146 5,325 6,769 8,592 29%

VIDEO CALLING

308 442 659 905 1,251 1,736 41%

ONLINE GAMING

49 68 95 133 187 290 43%

VOICE OVER IP (VOIP)

138 147 153 157 160 168 4%

OTHER 0 1 1 3 8 11 132%

12,528 17,867 24,477 32,973 43,772 58,214

Figure 9

YEAR

PET

ABY

TES

* Most relevant for the music industry

SOURCE: Cisco Visual Networking Index 2011

40 41

In 2009, there were 2.7 billion mobiles in use, with mobile calls accounting for approximately 125 million tonnes of CO2, which is just over 0.25 per cent of global emissions.

If/when we reach the penetration whereby there is one mobile handset per capita, we are getting close

to mobile calls accounting for one per cent of the global CO2 emissions. Additionally, as users convert from traditional handsets to smartphones, assuming everything else stays the same, CO2 emissions are only set to rise. Thus mobile handsets will soon consume more energy than the data centres did in 2009.

In his book How Bad Are Bananas?: The Carbon Footprint Of Everything, Mike Berners-Lee concludes that the carbon footprint of using a mobile phone is 47kg CO2 from a year’s typical usage of just under two minutes per day, while it mounts to 1,250 kg CO2 from a year’s usage at one hour per day, and 125 million tonnes of CO2 for global mobile usage per year.

‘A minute’s mobile-to-mobile chatter comes in at 57g, about the same as an apple, most of a banana or a very large gulp of beer. Three minutes has a similar impact to sending a small letter (written on recycled paper) by second-class post.

Mobile phones cause a fairly tiny slice of global emissions, but if you are a chatterbox using your mobile for an hour each day, the total adds up to more than 1 tonne CO2 per year – the equivalent of flying from London to New York, one way, in economy class.

Indeed, the footprint of your mobile phone use is overwhelmingly determined by the simple question of how often you use it. One estimate for the emissions caused by manufacturing the phone itself is just 16kg CO2, equivalent to (the energy consumed in producing) nearly 1kg of beef. If you include the power it consumes over two typical years (that’s about how long the average phone remains in use, even though most could probably last for 10 years) that figure rises to 22kg (of beef production).’

But the footprint of the energy required to transmit your calls across the network is about three times all of this combined, taking us to a best estimate of 94 kg CO2 over the life of the phone (or 47 kg per year). Broken down into data traffic, phone energy and manufacturing/shipping, provides the following energy share (figure 10).

76.33% DATA TRAFFIC

6.82% PHONE ENERGY

16.84% MANUFACTURING AND SHIPPING

CARBON FOOTPRINT

This data accords with the two studies referred to in 1.1.

Figure 10

ENERGY SHARE OF MOBILE

42 43

166,600 million tones CO2

(33.320% of global emission)

510 million tones CO2

(0.102% of global emission)

1,020 million tones CO2

(0.204% of global emission)

935 million tones CO2

(0.187% of global emission)

76,500 million tones CO2

(15.300% of global emission)

8,500 million tones CO2

(1.700% of global emission)

Another parameter (which is essential for the final carbon footprint figure) is the type of energy which is used. Figure 11 is from a study conducted by the World Nuclear Organisation42, for comparing the emissions from nuclear power with other power generating methods.

Wind, nuclear and hydro-electric power plants are clearly less emission-intensive and with nuclear still being seen as controversial, wind and hydro-electricity currently appear to be the most preferred, with the ICT sector increasingly looking to clean up its energy sourcing.

After a two-year campaign by Greenpeace43 against Facebook (and supported by millions of Facebook users), such was the pressure brought to bear on its coal-fired data centre in Oregon, USA that Facebook announced on 15 December 2011 that it would work together with Greenpeace to move to a renewable energy-powered data centre.

On clean energy investment, Greenpeace’s current report44 comments:

‘The International Energy Agency (IEA) warned in (Autumn) 2011 that unless a decisive shift is made to clean energy investment and away from high-carbon sources of energy, like coal, in the next five years (by 2017), the Earth will be locked into a disastrous cycle of unavoidable global warming. Electronic devices and the rapidly growing cloud that supports

our demand for greater online access are clearly a significant force in driving global energy demand.

‘While many brands are taking steps to manage and reduce pollution by increasing efficiency in their data centre operations, only a few companies have demonstrated a significant commitment to meeting their growing electricity needs from renewable sources.

‘This disconnect highlights the tremendous urgency in ensuring that these long-lasting investments in building the infrastructure to deliver the cloud are directed toward renewable sources of energy, and do not lock us in to our addiction to coal and other dirty sources of energy.’

Furthermore the same report states that:

‘The global telecoms sector is growing rapidly. In 2011, it is estimated that 6 billion people or 86.7 per cent of the entire global population have mobile telephone subscriptions. By the end of 2012, the number of mobile connected devices is expected to exceed the global population. Rapid growth in use of smart phones and broadband mobile connections mean mobile data traffic in 2011 was eight times the size of the entire Internet in 2000.’

While this forecast is significantly higher than Cisco’s 2011 report, the point therein, that the global telecoms sector is growing rapidly, remains the same.

42 World Nuclear Organisation43 Unfriend Coal 201044 How Clean Is Your Cloud 2012

GAS COMBINED

CYCLE

COAL

SOLAR

WIND

NUCLEAR

HYDRO

ENERGY SOURCE

ENERGY SOURCE EMISSIONS

Figure 11

44 45

APPENDIX 3FURTHER READING

DATA CENTRES IN THE US

WORLD ENERGY OUTLOOK 2010

US ENERGY OUTLOOK 2010

1 KWH = 3412 BTU

DATA CENTRES AND ENERGY CONSUMPTION (2011)

HP PROJECT MOONSHOT

INTERNET PENETRATION

CARBON FOOTPRINT - POWER GENERATION

CARBON FOOTPRINT – MOBILE PHONE

THE ENERGY AND CLIMATE CHANGE IMPACTS OF DIFFERENT MUSIC DELIVERY METHODS KOOMEY (2004) & TAYLOR AND KOOMEY (2008)

ACKNOWLEDGMENTS

Grateful thanks to Julie’s Bicycle for help and guidance in shaping this report

• The Norwegian Research Council for supporting the primary research project ‘Sustainable & Green Solutions For Online Media In Enhanced Networks’ in 2009 and to agree to include a task to address the potential energy consumption aspects of data traffic and data centres, in the absence of meaningful study by the wider ICT community.

• University of Agder, Norway, for hosting a presentation of initial findings from primary research studies at a music industry roundtable discussion in Kristiansand, 2010.

• Fraunhofer IDMT for assisting the project team in better-understanding the various P2P protocols and how they could be utilised in future services.

• Dr. Mohammad Ravanbakhsh at the University of Bergen Faculty of Mathematics and Natural Sciences, Department of Informatics) for assistance with finding appropriate reference material and for transforming technical and scientific R&D reports into a robust method to calculate energy consumption across various networks and applications.

THE AUTHOR WISHES TO THANK:

46 47

ABOUT THE AUTHOR

Dagfinn Bach Bach Technology President

Dagfinn Bach is President, R&D director and co-founder of Bach Technology AS, the company behind the MusicDNA format. Working for Western Norway Research Institute in the late 1980s Dagfinn led a cluster of pilot projects including development of the MP3 in music production and distribution, digitisation of music archives, and the creation of one of the first mixed-mode audio/multimedia CD-ROMs.

Dagfinn went on to become the initiator and coordinator of several important European Commission funded projects, and was appointed external expert and evaluator for the INFO2002 Multimedia Rights Management Systems call for proposals in 1998. Following this Dagfinn was hired as consultant for Nokia Ventures Organisation to conduct a feasibility study on mobile distribution of music in the Chinese mainland market.

Since founding Bach Technology in 2007, Dagfinn has overseen the development of the MusicDNA format. MusicDNA is a smart media extension that enables music fans to access the wide range of music-related content they want alongside the music itself – from lyrics, artwork and tour dates to blog posts, videos and twitter feeds all in one application.

MusicDNA allows content owners to create products that give back to the music fan that deeper experience they had when music came on a physical format. Launched at Midem in January 2010, MusicDNA has been hailed as the successor to the MP3 with support from across the value chain, including rights holders, distributors, digital service providers and retailers.

ABOUT MUSICTANK

Unique among the music business’ many and various interest bodies, MusicTank is a unique, neutral information hub for UK music business addressing change and innovation through informed debate, objective analysis and industry engagement.

Established in 2003 to inform and guide the future shape of the music business through engagement with industry, change and innovation, MusicTank continues to enjoy a growing and enviable reputation for its on-going and far-reaching programme of think tanks, conferences and events and has become the accepted provider of quality debate and analysis to the business. Bringing hot topics into sharp focus and helping pinpoint the opportunities created by disruptive technologies, it facilitates the circulation of innovative ideas, best practice and cutting-edge strategies for increased innovation and productivity. MusicTank is one of University of Westminster’s sector-based Knowledge and Business Development Networks.

The copyright in this report is held by University of Westminster. This material may not be copied or reproduced wholly or in part for any purpose (commercial or otherwise) except for permitting fair dealing under the Copyright, Designs and Patents Act 1988, without the prior written permission of its owners.

The copyright owners have used reasonable endeavours to identify the proprietors of third-party intellectual property included in this work. The authors would be grateful for notification of any material whose ownership has been misidentified herein, so that errors and omissions as to attribution may be corrected in future editions.

48 49

REPORT PARTNER

HEWLETT-PACKARD

Accelerate your innovation with HP

HP’s high-performance computing solutions are inspired by you and purpose-built for the complex problems that you’re working to solve today and tomorrow. We bring together the intelligence and automation of our HP ProLiant Gen8 servers, with some of the fastest processing power and largest deployments on the planet along with data centre innovations that pack more servers into less space and deliver greater performance per watt to give you a competitive edge.

So go out and solve some of the world’s big problems. We got your HPC data center.

• Smarter, easier deployment lets you adapt quickly to change and improve quality of service

• Faster scientific and research advancements with HP converged infrastructure that is purpose-built for scale

• Better density, efficiency and affordability to lower your total cost of ownership

REMAKE REMODEL: THE EVOLUTION OF THE RECORD LABELTony Wadsworth with Dr. Eamonn Forde, 2011

MusicTank’s much-anticipated fourth industry report was unveiled at the Great Escape Festival, Brighton UK on Fri 13th May, 2011. The report draws from the quarter century Tony has spent on the front lines of the record business – including the most tumultuous decade in its history. It defies much of the accepted wisdom and reveals the latest thinking on the evolution and future of the record company through conversations with some of the sector’s top executives.

“The document is definitely worth reading for anyone in this business...”(Emmanuel Legrand May 2011)

LET’S SELL RECORDED MUSIC! Sam Shemtob, 2009 Analysis of the relationships between the recordings business, ISPs, consumers and Government, in an attempt to help foster progress towards compelling legal alternatives to unlicensed file sharing.

“…a thoughtful study…It’s a great read on where that world finds itself right now – caught between lawsuits, the threat of government regulation, and a failing business model.” WIRED March 2009

MEET THE MILLENNIALS, Terry McBride, 2008

An authoritative account of fan influence over musical creation, exploitation and consumption.

“…a new, thought-provoking essay on the current face of the musicians’ industry.” Billboard May 2008

“…particularly engaging…a fast-paced, incredibly enthusiastic, positive report…it really ought to be bedtime reading for anyone who is attempting to generate success in this disparate, complicated industry.”

BEYOND THE SOUNDBYTES, Peter Jenner, 2006 Analysis drawn from MusicTank activities and ‘filtered’ through the author’s 40+ years of music business experience, detailing key industry issues and suggestions for their possible resolve.

“…getting the conversation started certainly seems like a good place to begin – and some of the ideas Jenner raises in ‘Beyond the Soundbytes’ are provocatively compelling” Stereophile.com November 2006

REMAKE REMODEL: ABOUT MUSIC TANKREMAKE REMODEL: ABOUT MUSIC TANK

REMAKE, REMODEL: THE EVOLUTION OF THE RECORD LABEL

50

musictank.co.uk

CONTACT DETAILS

MusicTankRoom E1.2University of WestminsterWatford Road, Harrow, Middlesex HA1 3TP

T +44 (0) 20 8357 7317E info@musictank.co.uk