Q&A: Dr Steve Hodgkinson, Ovum

(With internet usage now growing at rate of over 50% annually, there is an increasing focus on the energy requirements - and consequent environmental impact - of such a massive flow of information. Dr Steve Hodgkinson, research director at analysts Ovum, spoke with SSON about the difficulties of assessing this impact with any accuracy - and what organisations can do to minimise their own IT-related carbon footprints.)

SSON: Steve, an increasing amount of attention is now being paid to the impact internet usage is having on climate change. Many people may be astonished to learn some of the headline figures – notably, the fact that just one Google search generates anywhere between 0.2 and 7 grams of CO2! Can you explain how this is the case? How is all this extra carbon being generated? And why is there such a big range of estimates in that Google statistic?

Steve Hodgkinson: The science and methodology, not to mention the motivations, behind these calculations is at an early stage of evolution – so we can expect to see a wide divergence in estimates from different observers until some credible non-partisan measurement is done.

The motivation behind the estimates is an important consideration. Google’s motivation is to seek to minimise their perceived contribution to CO2 emissions, while other parties are seeking to over inflate estimates in order to sell services aimed at CO2 footprint measurement, energy efficiency tactics or carbon trading etc.

Where is the carbon generated? By the generation of electricity and its consumption throughout global ICT network – from the computer/laptop to the router/modem to the network equipment to the data centre servers and all the supporting infrastructure to keep it all running … transformers, cooling systems, CPUs, RAM, disk drives etc. etc.

Why such a range of estimates? Well, apart from the motivation of the observer, there is not yet a clear consensus on methodology. Where do you start and stop counting the watts and joules? Google’s numbers were, I think, based on the server processing workload once a search enquiry crosses their fence inbound and outbound (as it were). The research that was (incorrectly as it turned out) quoted in The Times to suggest 7gms was based on a more complete energy cost, including the desktop/laptop and telecoms network etc. Should you also ‘load’ the estimates with the full lifecycle energy cost of all of the components (i.e. the cost of their production, distribution, retailing and disposal/recycling), or just the marginal usage cost?

SSON: Do we know to any degree of certainty the carbon footprint of IT and IT-enabled activities globally? And how does this compare with other activities such as flying or agriculture – in other words, how serious a contributor to climate change is the computer?

SH: "2% of global CO2 emissions" seems to be the current view, with the most authoritative study that I am aware of being the United States EPA’s 2007 study*. This study concluded that data centres consumed 1.5% of all the electricity consumed in the USA, so ‘2%ish’ is plausible once ICT infrastructure outside of data centres is added to the mix … 2-3% seems a reasonable estimate to me.

The aviation industry is estimated to produce around 3%, so this seems to be being used as a comparator, just because it is a bit more tangible. It’s an odd comparison, however, given the substitution effect of digital communication vs. physical travel.

SSON: You’ve suggested that getting people to reduce their internet usage - in the same way that some environmentalists are encouraging people to cut back on flying, for example – might actually be counter-productive. Why?

SH: Well, if you are a global warming warrior obviously there is a theoretical argument to avoid any unnecessary activity that generates heat or consumes energy – and clearly using any form of ICT does both. Taking this extreme view would lead one to say that using computers and the internet generates CO2 and therefore their use should be minimised. Fair enough in theory, with the same logic as minimising air-conditioner use and unnecessary car journeys.

My comment though, is in reaction to the emerging hype around "evil energy-consuming internet data centres" and the impact of an individual’s internet use. The fact is that the world’s ICT infrastructure is currently very energy inefficient because it has been built over the past 10 years or so with little regard for energy consumption and CO2 emissions. The EPA report concluded that many data centres had power usage effectiveness (PUE) ratings of 2 – which means that half of the electricity that goes into the data centre is wasted in power conversions and cooling and so on before being applied to any useful computing work. This sort of wastage of energy pervades the world’s ICT infrastructure from the desktop/laptop right through to the data centre.

The issue is that the only way to solve this problem is to modernise hardware, software and management practices throughout the ICT infrastructure. There are incentives in place for the industry to do this because it is actually all about reducing costs and boosting profit. The companies involved, however, need robust revenues to make these modernisation investments… So it is actually important for people to keep using the internet.

Indeed, for most transactions the internet will be the least-CO2-footprint way of achieving a particular task, finding information, shopping, communicating, meeting etc., so we ought be seeking to boost internet usage rather then constraining it under bogus arguments about the impact of internet use on CO2 emissions.

In addition, the marginal impact of an individual person’s internet use is so small in the broader scale of the energy cost of the desktop/laptop device on the one hand and the demand-insensitive, largely fixed, energy cost of the data centres on the other. Data centres' energy consumption is not very proportional to demand … even if demand drops off to zero the CPUs, RAM, disk drives, routers, lighting, pumps, fans etc. etc. still consume energy.

So, rather than seeking to curtail demand, we should be advocating simple energy efficiency initiatives such as turning off devices when they are not being used and stimulating investment in the data centres to make them more efficient and more proportional in their energy consumption and to connect them to greener sources of electrons.

SSON: So what can organisations do to reduce the impact on the environment of their existing IT facilities? What, for instance, are the implications of adopting shared services?

SH: In truth it is just common sense. The first step is to measure and analyse energy consumption and seek to make it transparent – so that staff can voluntarily change behaviours that reduce energy use (a good example is the miraculous reduction in mobile phone costs that occurs when the bills are made transparent to users rather than hidden in a corporate accounts dept.)

Consolidation and rationalisation of facilities – into shared services for example – can be a viable way to both reduce the amount of computing work performed (by reducing duplication and nugatory effort) and also to create the opportunity for investment in new generation technology.

Start at the desktop and work backwards. LCD/Plasma monitors and laptops use less energy than desktops and CRT displays. Turn devices off automatically when they aren’t being used. Update heating and cooling systems in the data centre and only run them when required – perhaps allowing an increase in the operating temperature. Open the windows in the data centre!! Virtualise servers etc. to optimise asset utilisation. Explore strategies for making the energy consumption of the equipment more directly proportional to actual usage patterns. There are many strategies, and the leading ICT vendors are all focused on solving the various technical and operational problems … so that they can sell new hardware, software and services.

The other stream of options is sourcing electricity from ‘greener’ sources of supply – such as solar, wind, hydro or geothermal generation rather than brown coal. This depends very much on the availability of ‘green energy’ from electricity generation and retail utilities – and it is more expensive (and hence is another reason why the companies need to sustain robust demand). Other options include using co-generation plant onsite if there is a source of waste heat available, from other manufacturing processes perhaps.

SSON: How can organisations putting in new IT facilities (rather than updating existing ones) ensure that they are as "green" as possible – and can this be done without significant extra costs?

SH: It is a matter of designing the new facilities with energy consumption and heat production as explicit design criteria, and then sourcing the appropriate advice from the various service providers: architects, engineers, ICT infrastructure, workplace designers etc. It's not rocket science, but it just needs to be explicitly considered rather than treated as an afterthought.

SSON: How can firms looking to outsource their IT ensure that their providers keep emissions as low as possible?

SH: Make energy considerations an explicit part of your requirements, the supplier evaluation, and the contract.

SSON: Do you see the need for international regulation on this issue?

SH: There may be a need for some form of authorised basis of measurement and certification of energy efficiency (like the star ratings used for consumer appliances), with Environmental Protection Agencies being the logical administrators. I suspect that more regulation is inevitable, but my preference is for market solutions based around the rising awareness of consumers and the natural economic incentives on all parties to minimise the costs of the energy that must be purchased.