How can engineers make IT more sustainable? Part 1: It’s holistic

The digital sector generates 4% of global greenhouse emissions. Data centers and hardware are the main contributors to that impact. But all those machines — not to mention all software, websites, apps and more — run on code written by developers. This means their work’s impact is far from negligible. So how can engineers, and indeed anyone shaping tech today, reduce that impact?

Starting now, we’re going to be taking an in-depth look at the other side of green IT: the part we can all do something about. We’ll do this across four chapters, starting off with an overview. Enjoy!

What to know first: it’s holistic

Close to 4%, the digital sector’s greenhouse gas impact is almost double that of aviation. Though flying gets a lot of press due to its considerable emissions, only 2% of the global population actually flies. Whereas everyone has a computer, or a smartphone, or indeed several devices. This could explain why some experts consider 4% to be quite low. According to France’s national environmental agency, ADEME, this proportion could triple by 2050 if our habits don’t change.

As temperatures soar worldwide and carbon neutrality deadlines draw ever closer — we have to reduce our carbon footprints by 80% by 2050 if we are to reach the target of 1.5°C temperature rise set by 2015’s Paris Agreement — we could be panicked into throwing away our laptops and going to live on a desert island. The good news is, we don’t have to. Better still: the choices made by those working in tech can have a positive impact, right now.

What we must do, however, is look at all of the sources of our digital lives’ impact holistically, as they are all co-dependent. None can be taken in isolation. Then, we should measure them, and see what we can do to reduce each one. Those sources are:

• Data centers
• Hardware
• Code (software and websites/apps)

So let's dive into each one!

Data centers: the motors of digital life

No data centers, no cloud, no internet, no digital. We simply can’t live without these server-packed digital factories, whose efficiency is the key to their potentially huge impact. What level of impact are we talking about? It depends how you look at the situation.

Globally speaking, data centers represent 1% of energy-related greenhouse gas (GHG) emissions (IEA) and 1% of the world’s total electricity use in 2018 (IEA).

In France, where digital’s total GHG emissions are 2.5% (less than the global average), only 4% of those can be attributed to data centers’ energy usage. However, also according to the IEA (and Enerdata), data centers globally use almost as much electricity as countries like the UK:

• UK: 286 TWh
• Indonesia: 266 TWh
• Data centers: 200-250 TWh
• South Africa: 208 TWh
• Egypt: 153 TWh

TWh = terawatt hours, in 2020. Source: DW

It all depends, of course, on how many data centers are located in a given area: the UK is the European country which hosts the most data centers of its continent, for example. Still, as widely as these figures vary, it remains vital to dig in to how your cloud provider is contributing to that impact… and if it can be reduced.

The key factors of data center sustainability are:

1. Cooling
2. Energy sources
3. Servers

So let’s take a look at each of them.

1. Cooling a data center with air conditioning (AC), the most common method, can use up to 40% of a facility’s total energy consumption. That could be halved if data center clients accepted their servers being cooled at 27°C (a perfectly functional temperature) rather than the standard 22°C. As for the HFC gasses which often leak from AC systems, these are anything from 1000 to 9000 times more potent GHGs than CO2. Which is why reference work Project Drawdown has identified Refrigeration as the number one problem to solve in order to halt global warming.

Then there’s water. Microsoft’s facility in Middenmeer, Holland, was found to have consumed 84 million liters of water in 2021, at least four times more than what the company had promised. Why? Because cooling towers, the technique used in these cases, require immense amounts of water in order to be effective.

Discerning developers should as such look for cloud providers using alternatives to the above two methods. Free cooling, for example, uses nothing but outside air to keep servers cool; and adiabatic cooling lowers outside air temperatures by passing it through a humid membrane, with minimal energy and water expenditure.

2. Energy sources are also crucial here. Data centers can use 100% renewable energy, first and foremost. Failing that, users should privilege facilities in countries with the cleanest possible energy mix, i.e. that use as little carbon as possible. France is a good choice here, as the majority of its electricity comes from nuclear or hydroelectric power stations. Countries like Poland, meanwhile, are still highly reliant on coal. Not forgetting timing: running big compute jobs at times when a country’s electricity carbon mix is as low as possible can work wonders for the planet.

3. Servers, last but not least, are an essential factor in data center efficiency. Does your cloud provider throw away its servers after 3 to 4 years, the industry average? Or does it make them last as long as possible, thereby reducing hardware GHG emissions and e-waste? More on that below (remember, we told you all these factors are intertwined? 🤓)

Hardware: the elephant in the room

If you have to charge your smartphone every day, it can feel like it’s using a lot of electricity. But the fact is, over 80% of a modern mobile phone’s GHG emissions come from its manufacturing, not from its daily use. This is notably because of the rare metals they require, which need to be transported all over the globe. It as such makes sense that hardware accounts for more than three quarters of the digital sector’s GHG emissions (again, in France, but the proportions remain similar worldwide). As such, any green IT strategy must focus on devices as a n°1 priority.

Naturally, not all devices have the same manufacturing impact. That of servers — the ones that populate data centers — is just 15-30%, for example. But whatever the impact, given devices’ massive overall GHG footprint, the importance of maximizing their lifespans is obvious. To take a Scaleway example, the recent decision to refit 14,000 servers tripled their lifespan and saved the GHG equivalent of 5000 Western European households’ annual electricity consumption. And it’s not just GHGs: e-waste is the fastest-growing type of waste today.

So making hardware last as long as possible should be a priority for cloud providers, developers, and indeed all of us. Do you know if your cloud provider reuses its servers, or does it bin them after the average 3-4 year usage time? Do your preferred tech brands and retailers clearly display their products’ repairability index in stores? Do your favorite manufacturers do all they can to make their products easy to repair? Fairphone, which just released a new pair of repairable headphones, is a pioneer in this field. On the other hand, Apple only recently caved to over a decade of consumer pressure by making iPhone repair kits available.

Another key green IT hardware question for cloud providers and device manufacturers: are they using the most efficient components in their hardware, or rather the most powerful? Alas, the tech sector has traditionally leant towards the latter. Think Moore’s Law, which says that microchips’ power doubles every two years; then multiply that by the Jevons paradox, whereby gains in efficiency are supposed to lower resource consumption, but turn out to do the opposite. In other words, the faster we can go, the faster we do go, whether we actually need to or not. And with AI, this trend is only set to accelerate. One solution of many: chips like Ampere’s Altra Max series, for example, which deliver the same performance as x86 processors, whilst using 2.8x less energy.

As we’ll see in part 2, the above factors contribute to a device’s embodied carbon rating, or the total amount of carbon emitted by its creation, usage and destruction. So if, for example, a smartphone’s embodied carbon is 1000 kg for its entire lifecyle, it’s better to spread that impact out over three or more years, rather than 2.3 (the average time smartphones are used).

Code: the hidden culprit

Proportionately speaking, developers can’t have as much impact as data center operators or hardware manufacturers. Or can they? Whilst many aspects of this debate — such as whether some coding languages use more energy than others — are still totally open, some are already beyond discussion.

Did you know, for example, that infrastructure and software inefficiency count for over 50% of GHG emissions in data centers, according to Intel?

Or that a major culprit behind hardware obsolescence — and therefore e-waste — is code? Have you ever bought a new smartphone because your previous one couldn’t run the latest versions of your favorite app? This is because the code hasn’t been written for as many past versions of the mobile operating system as possible. So making code as backwards-compatible as it can be should be top of green IT-focused developers’ minds.

Then there’s the issue of efficiency. Some current trends are inescapable:

• The average size of a web page has increased by 191% over the past 10 years (source)… and that's on desktop. On mobile, it's +436%…
• Microsoft Office 2019 needs 171 times more memory than it did in 1998 (source)

Are websites 191% better today than in 2013? Is Office 171 times superior to its 1998 version? This is where the importance of clean, light and efficient code comes in. It’s not just about which language, or the number of lines used; the type of algorithms is also crucial.

SonarQube, the tool which evaluates code quality, is a green IT developer’s friend here, as is EcoCode, its sustainable equivalent; but again, more on that later. Measuring is key, as is trying alternatives, like building static websites, which only consume resources for the content a user requires to see at a given time, rather than loading a heavy CMS like WordPress every time.

Suffice to say, there are more and more examples of clever coding reducing energy consumption and improving cloud and hardware performance… and we’ll be sharing them with you soon.

So watch this space for part 2, when we’ll be focusing on how to measure cloud, hardware and software’s impact!

Thanks to the Green Software Foundation, and Octo’s Tristan Nitot, for several of the examples & notions in this article.

This blogpost is extracted from the Scaleway white paper "How can engineers make IT more sustainable?", which you can download in full for free here!