With heat waves gripping 3 continents and global temperatures reaching record highs this summer, perhaps we need to pay closer attention to ‘the cloud’ and computing’s environmental impact
Burning fossil fuels, like coal, oil, and gas, is the largest contributor to climate change, but while that fact is often exclusively associated with cars, air travel, and factory emissions, did you know digital technologies make up about 4% of carbon emissions globally? Or that energy consumption is rising 9% per year?
It’s easy to picture emissions from a factory or car but much less intuitive when it comes to someone developing software on their laptop. It’s a potentially huge blindspot for curtailing emissions, and the public sector needs to hold the tech industry more accountable for emissions produced by the infrastructure used to deploy software.
Computing’s environmental impact on our planet
That developer’s laptop doesn’t operate in a vacuum. It and the thousands of servers in the thousands of data centres making up the cloud are also big producers of carbon emissions, surpassing the emissions of over 22.2 million flights annually. Globally, the UK ranks third among countries in terms of the number of data centres, with 456 in 2022.
While the UK is generating more of their electricity from renewable sources than ever before, at 40% renewable in 2022, the government has admitted that their current “net-zero strategy will fail to cut greenhouse gas emissions enough,” according to the Financial Times.
Net-zero strategy will fail to cut greenhouse gas emissions enough
When these emissions are out of sight and out of mind, it’s easy to disregard their impact. However, when we look at how energy in data centres is consumed, it becomes clear that that impact is significant.
The cloud’s carbon footprint was estimated to be over 2% of global electricity production
According to research from Yale in 2018, the cloud’s carbon footprint was estimated to be over 2% of global electricity production, with more recent data suggesting it’s now 3%. To put this in perspective, a single data centre consumes the electricity of 50,000 homes. When we multiply that by the 456 data centres in the UK and 8,000 globally, the magnitude of the impact becomes staggering.
Shockingly, about 88% of that electricity isn’t even utilised for computational processes; it’s used to ensure the cloud stays up 24/7 via cooling and maintaining redundant fail-safes.
In short, ‘out of sight, out of mind’ is not a valid strategy to meet the UK Government’s ambitious net-zero targets. Under the Paris Climate Agreement, they have pledged to cut emissions by 68% by 2030, a mere 7 years away. This goal is meant to be a building block towards reaching climate neutrality by 2050, that is, reducing its greenhouse gas emissions by 100% from 1990 levels, in tandem with European Climate Law.
We need every tool at our disposal to cut emissions, and as it happens, one unlikely candidate worth considering is cloud-native container orchestration technologies – specifically Kubernetes.
Cloud-native can reduce environmental impact
Kubernetes, whose name originates from the Greek word for ‘helmsman’ or ‘pilot’, is an open-source container orchestration system. Like the captain of a ship, it’s in charge of managing, or orchestrating, containerised cloud applications, always ensuring the right computing, network, storage, and configuration.
It allows organisations to spin up, terminate, update, and scale-up applications with better resilience, governance, security, visibility, and lower operational costs.
However, when viewed from an environmental perspective, Kubernetes boasts less-obvious advantages. It automatically and intelligently grows or shrinks computer power based on what’s needed, preventing resource waste. This way, you’re never leaving resources idle or using too many.
AI and ML: Future tech and our changing environment
Given the AI craze we’re going through, this is going to be significant because AI/ML workloads are very taxing on the hardware. And as mentioned earlier, an overwhelming majority of electricity consumed in data centres is used for cooling that hardware, with up to 40% spent on cooling alone. With AI’s growing popularity, that number is only going to get worse.
Kubernetes could play a significant role in alleviating this because while it’s not specifically designed to be a ‘sustainability tool’, it is designed to manage and avoid hardware redundancies. It could optimise and streamline those AI workloads (or any workloads) at scale.
Even something as basic as software testing environments can see lower footprints by using Kubernetes. The old model of ‘un-containerised’ virtual machines would force companies to leave a testing environment running all the time – even when they’re not testing. In contrast, Kubernetes lets you scale a testing environment up and down as needed. This is important because more time spent developing or testing equals a bigger carbon footprint.
It matters ‘when’ you adopt cloud-native tech
It’s no secret that one of the reasons the UK public sector has struggled to digitally transform is the slow pace of adopting new technologies.
This isn’t just a problem for efficiency and innovation. Your environmental impact is affected by when you adopt Kubernetes. At my recent Kubecon presentation, Scale Down Your Environmental Impact, my colleague Zinnia Gibson and I demonstrated this by asking the audience to partake in a bit of a thought experiment:
Imagine two large-scale companies in the game development space with the same product. Company A uses Kubernetes, but Company B, which intends to use it down the road, still uses virtual machines.
The company using Kubernetes, because of its ‘autoscaling’ capabilities, automatically uses fewer resources from data centres by the nature of its infrastructure’s design, whereas the alternative leaves too much room for mismanaged resources at a large scale.
It’s also worth considering that any public sector organisation releasing an application that sees its user base rise dramatically in numbers will inevitably need to scale that application up. If they haven’t already integrated Kubernetes into their workload, they now need to worry about training and infrastructure. More time and energy spent on computing means more emissions.
The fight for sustainability will not be solved by one single solution
Of course, let’s be clear: none of this is to say that Kubernetes is ‘the one thing’ that’s going to defeat climate change. As one of my favourite voices in the sustainability space, Shelbi Orme says: “You cannot do all the good that the world needs, but the world needs all the good you can do.”
The fight for sustainability will not be solved by one single solution, person or industry; we all have a role to play. The tech industry’s potential to make a sizable impact cannot be ignored, and the public sector needs to think about computing the same way. Kubernetes is just one arguably effective way to start clamping down on unnecessary computing inefficiency and, in turn, help reduce computing’s environmental impact and carbon emissions.
The best part is that IT teams in the public sector can already access free upstream projects to decrease development time, lower cost, and lower environmental impact. There are many alternatives, but here are some favourites:
- GreenFrame: an open-source tool which measures and reduces your website’s CO2 emissions by detecting carbon leaks.
- Prometheus: not explicitly designed to track emissions but can help obtain metrics so you know when resource loss is occurring.
- Microsoft Emissions Impact Dashboard: a tool explicitly focused on showing carbon emissions for Azure cloud usage.
Any responsible organisation will look to renewable energy efficiency improvements to lower their carbon footprint (like making their offices more environmentally friendly, carbon sequestration, etc.). But with the climate outlook looking this dire, we have an obligation to use as many tools as we can. Kubernetes should be part of that mix.
This piece was written and provided by Mary Karroqe, Software Engineer at D2iQ
