The growth in data volume is tangible and there is indisputable evidence (e.g. Japanese internet exchange) that the CAGR of data volume is currently around the 40 percent mark. Vodafone has experienced 69% data growth in 2010 and, as we move into the world of “the internet of things” where billions of sensors and devices will enable such things as the smart-grid, the growth in data generation will accelerate even faster.
Enabling that growth is the roll-out of fast and super-fast broadband connections for both fixed and mobile devices. The UK government target is access to at least 2Mbps for all, even in the most remote rural areas, although the EU has a far more ambitious target of 30Mbps for each household. Verizon (using 4G LTE technology) delivers speeds of 6-12Mbps (with peak performance of 150Mbps) and Ericsson is currently demonstrating 1Gbps in mobile trials.
There are many anecdotal pieces to illustrate the growth:
- Twenty average domestic homes with FTTH (fiber to the home) can generate more internet traffic that the whole of the internet backbone carried in 1995 (Source, CISCO)
- More digital video volume is uploaded to YouTube every month than the digitized equivalent of standard definition terrestrial TV broadcasting 24×7 for 180 years (2,160x the volume)
Data-growth predictions vary by source but some examples are:
- CISCO 2600% by 2015 and 32% CAGR
- Nokia Siemens Networks 2500% by 2015 and planning for 1,000x capacity 2010-20
- IDC 4400% growth between 2009-20
- Gartner 1600% between 2005-25
Whatever growth rate we base a plan upon the roll out of fast broadband services will, no doubt, encourage more traffic volume, and volume that will be dominated by high-definition video.
Current power consumption of ITC
Data is carried over fiber, directed by routers and distributed and managed by data-centers. As the network expands in capacity and speed, the ease and convenience of generating and transmitting ever increasing quantities will generate more traffic.
Such transport and logistics of data consumes power and, depending upon which research you prefer to take, the UK ITC demand consumes between 2-3% of the national power generation. Latest reports for the USA put this figure at 6% – which, probably not coincidentally, is actually in-line with the c40% CAGR in data volume acting on the 2% power-consumption prediction of 2008 by the US government body EPRI. We are talking here of ‘ITC’, not just data-centers in isolation – something that is nearly always confused with overall ITC consumption.
As the futurologist Raymond Kurzviel suggested, if you only look backwards a few years even an exponential growth rate appears to be a straight line, and, to his point, we are in danger of ignoring the accelerating nature of the growth rate itself. So, whatever figure you wish to start with, be it 2%, 3% or even 6%, we will quickly reach a much larger number at 40% CAGR where “quickly” is measured in less than 5-year planning cycles.
As an assumption, we can take a proportion of the overall ITC predicted power demand to represent just that part that will handle the growth enabled by fast-broadband. The growth will impact on routers, data-centers and mass-storage arrays and we can safely assume that at least 40% of the current power consumption is concentrated in this part of the IT infrastructure. Therefore we will, in the later calculations use 0.8% (from the lower 2% UK prediction), 1.2% (from the higher 3% UK prediction) and 2.4% (from the US derived 6%).
ITC as a low-carbon enabler
There is no doubt that ITC can enable low-carbon behavior, e.g. home-working, video conferencing and reduction in travel, particularly by car. However we must accept that whilst superfast broadband is essential for business applications and actually stimulates GDP the same cannot be said for domestic downloading of HD-video, HD on-line gaming and many other social-networking applications. It is clear that a political aspiration may be to offer citizens the availability of fast broadband but if they “all” avail themselves and use the full capability, at what cost will this be in terms of power consumption and carbon-footprint?
It is interesting to contemplate the carbon-footprint difference between sending a hard drive (full of data, e.g. 2TB or 2,000GB – purchased for less than £200) by a document/parcel carrier with an integrated logistics infrastructure and sending the same data volume over the internet. Access to a super-fast connection (at both ends) means that the courier (with our inherent ‘now’ culture) would rarely be the transport method of choice, even though for very high data volume it would often be the lowest carbon solution.
Cautionary Note: We should remind ourselves that whilst ITC is an essential part of business and can, and does, enable low-carbon solutions/behaviour, internet traffic volume is dominated by content (particularly digital images and video) that has questionable social value, e.g. a few years ago an international long distance data carrier estimated that 60% of their long-haul data volume was pornography.
We need to consider if the growth in ITC power consumption will be offset by low-carbon enablement using ITC. It is doubtful, and certainly not obvious, if this will be the case.
The increased efficiency curve of the current technology generation
ITC hardware has been closely following Moore’s Law for four decades – the doubling of compute performance (actually transistors per chip but the effect is the same) every 2 years. This is predicted to last a further 5 years before the physical limits (at the one-atom width limit!) of how many transistors can be concentrated on a single microprocessor is reached. At the same time the power consumption ‘per unit of computation’ has improved in a similar way albeit at a slower rate. If we take a conservative view and assume that a doubling of capacity per Watt every three years applies to the power consumption for the next ten years, we can then compare the increase in data volume with the improvement in power consumption and draw a conclusion on the future power demand based upon a ‘starting’ value of 0.8/1.2/2.4% of the national grid capacity of 2010.
Power predictions for a ‘no natural limit’ internet
Parallels can be drawn between nature and the growth of internet traffic. The spread, and the speed of that spread, of internet access and the increasing bandwidth has resulted in a virus-like growth of data-generation and transport. In nature the spread of such a virus (as an epidemic) always has a natural limit and the epidemic ‘dies out’. It could be concluded that the generation of data on the internet has ‘no natural limit’ but power, and the ever increasing cost of that power, may be the mechanism that limits growth. In this respect the target to allow every citizen access to fast-broadband will be a waste of resource if that citizen can’t pay for a service that is dominated by the cost of power and the service itself becomes unaffordable within normal household budgets?
The following graph shows the relationship between starting value (how much power ITC may consume today), growth based on 40% CAGR in data-volume, efficiency improvement of the current paradigm in technology of 25% CAGR and the proportion of the grid power consumed by ITC for the next eight years.
Note that ‘1.00’ is the entire national electrical generation output of the UK in 2010.
Clearly, almost regardless of the starting value, the growth in ITC is unsustainable (when ITC consumes 10% of the national grid capacity) past 2015 even at the lowest 2010 starting value without a paradigm shift in technology and network efficiency. The improvement will have to be in 3-4 orders (1,000 to 10,000 times better than existing) rather than the present 25% CAGR.
The roll-out of superfast broadband has only just begun in the UK and a slow rate of take up will postpone the effect shown in the graph above. Without a paradigm shift in efficiency ITC power demand could rise by a factor of 10 by 2015 in total opposition to a plan for a low-carbon future. The question is this: This is clearly not a sustainable infrastructure plan but has the potential consequence, even in a small part, been taken into account by the government in setting the targets for unrestricted fast and superfast-broadband for all, everywhere and at any time?