We Need to Decarbonize Software

“There’s an already existing large segment of the software-development ecosystem that cares about this space—they just haven’t known what to do,” says Asim Hussain, chairperson and executive director of the Green Software Foundation and former director of green software and ecosystems at Intel.

What to do, according to Hussain, falls under three main pillars: energy efficiency, or using less energy; hardware efficiency, or using fewer physical resources; and carbon-aware computing, or using energy more intelligently. Carbon-aware computing, Hussain adds, is about doing more with your applications during the periods when the electricity comes from clean or low-carbon sources—such as when wind and solar power are available—and doing less when it doesn’t.

The Case for Sustainable Software

So why should programmers care about making their software sustainable? For one, green software is efficient software, allowing coders to cultivate faster, higher-quality systems, says Kaspar Kinsiveer, a team lead and sustainable-software strategist at the software-development firm Helmes.

These efficient systems could also mean lower costs for companies. “One of the main misconceptions about green software is that you have to do something extra, and it will cost extra,” Kinsiveer says. “It doesn’t cost extra—you just have to do things right.”

Other motivating factors, especially on the business side of software, are the upcoming legislation and regulations related to sustainability. In the European Union, for instance, the Corporate Sustainability Reporting Directive requires companies to report more on their environmental footprint, energy usage, and emissions, including the emissions related to the use of their products.

Yet other developers may be motivated by the climate crisis itself, wanting to play their part in fostering a habitable planet for the coming generations. And software engineers have tremendous influence on the actual purpose and emissions of what they build.

“It’s not just lines of code. Those lines have an impact on human beings,” says June Sallou, a postdoctoral researcher specializing in sustainable-software engineering at the Delft University of Technology, in the Netherlands. Because of AI’s societal impact in particular, she adds, developers have a responsibility to ensure that what they’re creating isn’t damaging the environment.

Building Greener Websites and Apps

The makers of COP28’s website could have taken a page from directories like Lowwwcarbon, which highlights examples of existing low-carbon websites. The company website of the Netherlands-based Web design and branding firm Tijgerbrood, for instance, emits less than 0.1 grams of carbon per page view.

Creating sustainable websites like Tijgerbrood’s is a team effort that involves different roles—from business analysts who define software requirements to designers, architects, and those in charge of operations—and includes green practices that can be applied at each stage of the software-development process.

First, analysts will have to consider if the feature, app, or software they’re designing should even be developed in the first place. Tech is often about creating the next new thing, but making software sustainable also entails decisions on what not to build, and that may require a shift in mind-set.

The design stage is all about choosing efficient algorithms and architectures. “Think about sustainability before going into the solution—and not after,” says Chiara Lanza, a researcher at the Sustainable AI unit of the Centre Tecnològic de Telecomunicacions de Catalunya, in Barcelona.

During the development stage, programmers need to focus on optimizing code. “We need the overall amount of energy we’re using to run software to go down. Some of that will come from writing [code] efficiently,” says Hannah Smith, a sustainable digital tech consultant and director of operations at the Green Web Foundation.

Tijgerbrood’s website optimized the company’s code by using low-resolution images and modern image formats, loading animations only when a user scrolls them into view, and removing unnecessary code. These techniques help speed up data transfer, loading, and processing on a user’s device. The website also uses minimal JavaScript. “When a user loads a website [with] a lot of JavaScript, it causes them to use a lot more energy on their own device because their device is having to do all the work of reading the JavaScript and running [it],” explains Smith.

When it comes to operations, one of the most impactful actions you can take is to select a sustainable Web hosting or cloud-computing provider. The Green Web Foundation has a tool to check if your website runs on green energy, as well as a directory of hosting providers powered by renewable energy. You can also ask your hosting provider if you can scale how your software runs in the cloud so that peak usage is powered by green energy or pause or switch off certain services during nonpeak hours.

AI the Green Way

Programmers can apply green software strategies when developing AI as well. Trimming training data is one of the major ways to make AI systems greener. Starting with data collection and preprocessing, it’s worth thinking about how much data is really needed to do the job. It may pay to clean the dataset to remove unnecessary data, or select only a subset of the dataset for training.

“The larger your dataset is, the more time and computation it will take for the algorithm to go through all the data,” hence using up more energy, says Sallou.

For instance, in a study of six different AI algorithms that detect SMS spam messages, Sallou and her colleagues found that the random forest algorithm, which combines the output of a collection of decision trees to make a prediction, was the most energy-greedy algorithm. But reducing the size of the training dataset to 20 percent—only 1,000 data points out of 5,000—dropped the energy consumption of training by nearly 75 percent, with only a 0.06 percent loss in accuracy.

Choosing a greener algorithm could also save carbon. Tools like CodeCarbon and ML CO₂ Impact can help make the choice by estimating the energy usage and carbon footprint of training different AI models.

Tools for Measuring Software’s Carbon Footprint

To write green code, developers need a way of measuring the actual carbon emissions across a system’s entire life cycle. It’s a complex feat, given the myriad processes involved. If we take AI as an example, its life cycle encompasses raw material extraction, materials manufacturing, hardware manufacturing, model training, model deployment, and disposal—and not all of these stages have available data.

“We don’t understand huge parts of the ecosystem at the moment, and access to reliable data is tough,” Smith says. The biggest need, she adds, is “open data that we can rely on and trust” from big tech data-center operators and cloud providers like Amazon, Google, and Microsoft.

Until that data surfaces, a more practical approach would be to measure how much power software consumes. “Just knowing the energy consumption of running a piece of software can impact how software engineers can improve the code,” Sallou says.

Developers themselves are heeding the call for more measurement, and they’re building tools to meet this demand. The W3C’s Sustainable Web Design Community Group, for instance, plans to provide a test suite to measure the impacts of implementing its Web sustainability guidelines. Similarly, the Green Software Foundation wrote a specification to calculate the carbon intensity of software systems. For accurate measurements, Lanza suggests isolating the hardware in which a system runs from any other operations and to avoid running any other programs that could influence measurements.

Other tools developers can use to measure the impact of green software engineering practices include dashboards that give an overview of the estimated carbon emissions associated with cloud workloads, such as the AWS Customer Carbon Footprint Tool and Microsoft’s Azure Emissions Impact Dashboard; energy profilers or power monitors like Intel’s Performance Counter Monitor; and tools that help calculate the carbon footprint of websites, such as Ecograder, Firefox Profiler, and Website Carbon Calculator.