The IT industry needs to work on efficiency to curb its unrelenting demand for electricity – and even if it does, by 2020 it will be responsible for four per cent of the world's greenhouse gas output, according to an Alcatel-Lucent-funded study.
The assessment, conducted by BIO Intelligence Services, suggests that the four per cent outcome is something of a best case. Its sobering – if unlikely – assessment is that without efficiency measures, the IT and communications sectors would generate nearly 330 gigatonnes of CO2 emissions by 2020. With efficiency measures, output would instead be 1.43 gigatonnes of CO2 - a CAGR of six per cent, which is well below adoption of IT and the rate of growth of Internet traffic.
More ReadingGreen supercomputer pioneer to model micro dronesCisco joins $8m gang, shovels cash into ex-NASA OpenStack cloud upstartWorld's 'most green' supercomputer in red-hot battle between Intel, NvidiaItalian 'Eurora' supercomputer pushes the green envelopeApple falls as rivals shake green tech tree
The reason The Register treats the worst-case as unlikely is that, as the report points out, 330 Gt worth of emissions would have the tech sector generating seven times the global greenhouse emission target for 2020 – it's more than ten times the 2010 emissions from energy consumption as documented at the US Energy Information Administration here.
In other words, the tech sector has to find ways to curb its consumption, if for no other reason than continuing to follow its current trajectory would make the availability of energy, not just emissions, a growth constraint.
The worst case is based on a high-end network traffic growth scenario by Bell Labs, in which 2020's total global telecommunications traffic would reach 400 times its level in 2007, and an absolute growth in power per user to more than 100 watts.
If the study is correct, achieving the necessary efficiencies will be a challenge: “The improvements that are possible in current technologies are expected to slow and then stop by the year 2017”, the report states – which means new technologies and efficiencies will be needed.
It's not all gloom, however: if the tech sector can stay within the “four per cent of greenhouse emissions” target, the study's authors believe IT and communications will ultimately save more emissions than they contribute.
Not all of the energy efficiencies need new technologies, the paper states. For example, infrastructure sharing – anathema to many mobile carriers, and in fact only common under either financial or regulatory pressure – is an obvious way to save power.
Cloud computing – also a form of infrastructure sharing – is also treated favourably by the study, which says it can reduce emissions by up to 90 per cent for small (sub-100 seat) deployments, falling to a best-case of 60 per cent for deployments in the 10,000-seat range.
Looking at the numbers
However, the key question, it appears to The Register, revolves around traffic growth predictions, since these are a fundamental variable in the study: how realistic is it to predict that 2020 network traffic could be 400 times 2007 traffic?
If I use Cisco's Visual Networking Index as a benchmark, the Bell Labs traffic growth forecast appears to be very optimistic. The monthly traffic recorded for 2007 in the company's 2008 VNI was around 6,500 PB per month, and the current forecast is for global IP traffic to pass 106,000 PB per month by 2016. Averaging the CAGR from 2007 to 2016, I get a value of 37.12 per cent.
If growth from 2016 to 2020 maintained that 37.12 per cent CAGR, 2020 traffic would reach 378,000 PB per month – around 59 times the 2007 traffic, far short of the 400-fold increase that the Alcatel-Lucent / BIO Intelligence Services predicts as the maximum growth rate.
However, the study provides a metric that can be applied back to the Cisco-based prediction: CO2 emission on a per-byte basis. The high-end forecast is for traffic to reach around 6.2x1010 Terabytes per year (thanks for mixing units there, authors). The short version is that on the same power curve, if traffic reached my extended Cisco VNI-based forecast, emissions would reach 42 Gigatonnes rather than 330, without efficiency strategies.
That's still a considerable emission growth, and still well ahead of what's desirable, since the global target is for just 45 Gigatonnes of emissions from electricity consumption.
The report can be found here. ®