This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies.Read our private policy

Energy use is being kept under control by improvements in chips. But data center deployment will drive consumption, so we’ll have to keep an eye on that.
09

Energy efficiency in ICT is cause for optimism

Progress faster than expected

By Anders Andrae, Huawei Technologies Sweden

Several years ago, Anders Andrae, a researcher at Huawei in Sweden, published an article about the growing volume of carbon emitted by the ICT industry. He looked at two main factors: how much global data traffic would grow; and how levels of energy-efficiency in the industry would change.

What’s happened since you wrote that article?
The first factor I looked at was global IP data traffic. From 2017 until now, it has increased at a rapid clip: about 26% per year, based on the volume of device shipments and the number of users connected to the internet. A growing number of internet users, combined with higher-bandwidth networks, has caused this steady rise.

The second factor, energy-efficiency, is a bit more encouraging. Gains in energy-efficiency at the chip level have arrived faster than expected, and there may already be chips that perform better than what had been predicted for 2030. Content transfer – moving bits across wireless and fiber-optic networks – is also improving with each generation of equipment and network design. Energy-efficient chips help decrease the power used in the computation/software share of ICT.

One of the take-aways from your article was that 5G was more energy-efficient, so therefore, the faster 5G was rolled out, the better off the planet would be. What’s happening there?
Global 5G roll-out has proceeded in line with expectations. Uptake is much higher in certain regions, such as China, the US, and South Korea, and much lower in others, including Europe and Africa.

According to my calculations, by the end of this year, 5G connections will make up about 16% of the world’s global wireless mobile and voice traffic. By the end of next year, it will have risen to about 22%. By year-end 2030, the figure will be somewhere around 94%.

In 2017 you estimated that by 2025, ICT could consume 20% of the world’s available electricity. How does this prediction look today?
Well, 20% looks way too high. Instead, ICT looks set to continue consuming about 10% of the world’s electricity through 2025. But the share of ICT production may increase as 2025 gets closer. That’s due to the rise of the Internet of Things, plus server-related demand not foreseen in 2017. Overall energy use is being kept under control by energy-efficiency improvements in chips. But data center deployment will drive consumption, so we’ll have to keep an eye on that.

What do you think of the notion of a “carbon handprint”?
The potential is great. From what I have seen, handprint calculations are well defined, simple enough, and business-oriented; and they are accepted by academia. But we need more examples of handprints calculated from the bottom up. That means a specific solution – for example, Huawei’s cloud storage – is analyzed using the handprint method with a particular customer. We should perhaps not try again to estimate the overall handprint of cloud computing on industry as a whole, as that would be a repetition of earlier work.

Are carbon handprints difficult to calculate?
Yes, they're complex. The global handprint of ICT has already been estimated, but complex individual cases are lacking. 

So far, the typical bottom-up handprint case studies in ICT are rather simple. For example, you can look at the use of liquid cooling vs. normal cooling in a cellular base station. That generates a handprint, because some emissions are prevented by the liquid cooling. But beyond that, there are no further discussions on exactly where the handprint is reducing emissions, and what the side-effects of that are. The handprint calculation needs to be refined (in standardization and beyond) to better quantify avoided emissions and rebound effects.

What’s a rebound effect?
When you implement new technology that changes some economic factor – such as cost, or profit – that may change the behavior of industry players, businesses, and consumers. Higher profits, for example, may be saved or spent in different ways. Those second-order changes are known as rebound effects. They’re hard to quantify, but we’re working on it.

Anything that makes you hopeful about the carbon future?
Semiconductors have become more energy-efficient than expected that's the major development of the last five years. New research in that area is very promising.

Another interesting trend is the prospect of low-, or even zero-emission data centers. These facilities can make use of renewable energy as a power source, optimize the re-use of wasted heat, and use free cooling where possible by siting data centers in colder parts of the world. AI, in conjunction with heat sensors, can also pinpoint the precise areas where more cooling is needed.

By contrast, traditional data centers are less efficient, and, given the rapid overall growth in data traffic and data center construction, could cause global ICT power consumption to increase significantly. Upgrading data centers to make them greener could really make a difference in the ICT industry’s carbon footprint in the coming years.


Contact us! transform@huawei.com