Green software

Our initial examination of Green software in 2023 has put this methodology firmly into the Adopt ring. Green software is an umbrella term for the various initiatives focused on the greenhouse gas emissions of software. With its focus on actually reducing the carbon footprint, we'd welcomed a mature attempt at addressing the environmental impact of the Tech industry. We had decided to follow the framework of the Green Software Foundation (abbreviated as GSF), as we'd thought they had avoided the usual greenwashing common to such industry-backed initiatives.

Since then, we did our own research at Kiwee and described it in our blog post titled Green software: Why efficiency and clean energy won’t cut it. We discovered that, upon a closer look, the understanding of our predicament with respect to climate change and environmental degradation, and the framework for addressing them proposed by the Green Software Foundation are too narrowly focused on two aspects: making software more energy-efficient, and shifting the energy sources needed to run it from fossil fuels to renewables.

Carbon tunnel vision

What we saw lacking in GSF's approach to Green software were several key considerations:

1. Software does not just need energy to run. It needs the hardware, made out of materials extracted out of the environment, which uses resources during its production, and is then disposed into the environment at the end of its lifecycle.
2. Renewable energy is not a get out of jail free card. Firstly, it has limitations that fossil fuels don't have to contend with (e.g. limited dispatchability). Secondly, just like the hardware, its infrastructure is still based on resource extraction, and there's now evidence that we don't have enough mineable materials to replace even a fraction of our fossil fuel consumption with renewables, let alone the question of how damaging it would be to actually mine them.
3. Improvements in software efficiency and the production cost of renewables sound like good news, but historically, both backfire and lead to higher energy and resource consumption. This is known as the Jevons paradox, or the principle of induced demand—whenever there's an efficiency gain, we do our best to consume more of the resource. Historian Jean Baptiste-Fressoz points out a related phenomenon with regards to energy transition—going so far as to state there never has been such a thing, as every new energy source simply adds to, and even reinforces, the energy usage from other sources.

What has changed in the last three years? It seems that the Green Software Foundation has recognized that greenhouse gas emissions are only a part of the story. They now include overall energy and water consumption and the hardware lifecycle in their definition of Green software. We might say they addressed point 1 from our list. What about point 2 about renewables? Unfortunately, the Foundation still treats them as a magic bullet—we didn't find any mentioning of their environmental impacts, or a discussion on the feasibility of running any sizeable chunk of computation on renewable energy. The Jevons paradox is also largely ignored in GSF's framework, even though they're definitely aware of it¹.

What GSF is focused at? The Foundation's main priority is coming up with ways to measure software carbon intensity. The hope is that once we can measure the carbon footprint, we can also reduce it. Firstly, carbon intensity is not the same as carbon footprint. "Intensity" here means emissions per usage unit (page visit, API request, etc.), thus completely avoiding the issue of rising consumption and induced demand. Secondly, even leaving all the larger-scope material barriers we mention above, there is one crucial problem, and that is incentive. Measurement is an important tool if you want to achieve an improvement, but there's just no evidence the Tech industry is even interested in reducing its overall carbon footprint, let alone all the other environmental considerations.

The elephant in the room

You might wonder why we don't mention AI here. While the AI boom has caused heated discussions about energy and water consumption, and the large-scale build-out of data centers has caused a lot of controversy, we believe it is fundamentally a continuation (and intensification) of the same structural issues outlined above.

Nevertheless, this leads to our final question and the actual elephant in the room: What do we actually want to reduce and compared to what baseline? As we demonstrated, carbon emissions from energy usage is but one component in a complex web of interrelated issues. It is relatively straightforward, though not necessarily easy, to measure it—at the end of the day, it's the amount of CO2 (its equivalent for other greenhouse gases) released to the atmosphere. But that doesn't mean other aspects (resource extraction impacts, feasibility of renewables, economic incentives) are less important to address.

Moreover, trying to reduce any metric only makes sense if there's a set baseline. If we don't have it, we'll end up with a scenario where we may constantly achieve 100x reductions in carbon intensity for a piece of software (e.g. an AI model), yet still increase the overall footprint if each new generation of said software has a different, higher baseline with which we measure².

Closing thoughts

Our own understanding of what it means to call software "Green" has been evolving over the years. We do see a more elaborate discussion on Green software coming from the industry, yet we believe it is still not mature and honest enough, taking into account the sheer urgency and scale of the problem. We focused on the Green Software Foundation because it has the greatest recognition in the Green software space, and is backed by leading Tech companies and reputable universities. It doesn't mean there are no alternative proposals. The World Wide Web Consortium (W3C) is actively working on the draft of its Web Sustainability Guidelines. We are seeing important considerations raised there, and the overall quality of communication feels more to-the-point than that of the Green Software Foundation.

However, we have decided to move the Green software item on our tech radar into the Assess ring—not because we believe it's not a topic worth addressing. On the contrary, we must put much greater effort in addressing climate change and ecosystems collapse, but the level of quality and actionable proposals that currently stem from the Green software milieu is just not substantial enough, and may miss the root causes entirely.

---

Footnotes

1. At least one on GSF's website statement mentions it in passing: "developers should ensure their code is efficient while keeping Jevons’ Paradox in mind". ↩

2. In conservation biology, this idea is known as the shifting baseline syndrome. ↩

Green software is software designed to reduce the carbon emissions produced during operation. There is no clear definition of green software, yet we follow the one created by the Green Software Foundation. The philosophy of green software is to reduce the carbon footprint, not to offset it.

Our commitment to green web applications stems from profound environmental concerns. Social responsibility and dedication to care about the natural environment are among Kiwee's core values.

There is a correlation between website performance and energy consumption. The fewer resources a web application uses, the lower its environmental impact is. Improved website performance is a win-win situation for the business and the environment. Users and search engines will appreciate faster loading of web applications. The companies will value lower server infrastructure bills.

How to measure carbon emissions generated by a website?

You can't improve what you can't measure. Accurate measurement of the carbon footprint of a single-page request is challenging. It requires data on application-level metrics and infrastructure. Most importantly, the environmental impact depends on the data center's use of renewable energy.

However, a good starting point is the website carbon calculator, which gives a result based on a formula that takes the payload size and constants defined for the estimated global and regional internet annual traffic and energy consumption.

In the future, we may expect websites to follow energy efficiency regulations similar to the European Accessibility Act or the Americans with Disabilities Act (ADA) regarding website accessibility.

How to make the website greener?

When making a website more environmentally friendly, it is important to address three areas: front-end, back-end, and infrastructure.

Green front-end Before any changes are made, we recommend running Google Lighthouse or a similar tool to get a point of reference.

The base steps are:

• Use images optimized for the web in next-gen formats.
• Avoid loading too large images.
• Remove Unused JavaScript and CSS code.
• Simplify and trim complex and heavy HTML structure.

We highly recommend applying the principles of green user experience (green UX) when working on new and existing projects.

Green back-end

• Optimize code for lower computational complexity.
• If feasible, choose modern languages like Go or Rust.
• Implement continuous performance testing.

Green infrastructure Cloud Carbon Footprint is an open-source tool to measure and monitor the carbon footprint of your cloud infrastructure for further optimization. Check if your website runs on green energy.

We recommend making the software more environmentally sustainable in your new and existing projects.

Want to explore this further? Read our blog post on why efficiency and clean energy alone won't be enough to make software truly sustainable.