As COP26 continues this week, there are signs that climate change is finally being taken seriously. Organisations worldwide are being put under more pressure to at least measure their carbon footprint, and then take steps to reduce it.
Currently, though, carbon-reducing actions that are publicly visible get prioritised. This is despite other less visible actions having a far greater carbon-reducing impact – including IT. One CTO recently said to me: “It’s crazy that we are going to so much inconvenience to reduce the carbon footprint of how we travel – because when you look at the numbers, by far our biggest source of carbon emissions is buried inside our IT infrastructure’s power consumption. Cutting waste there would be less inconvenient while delivering a greater impact.”
According to a 2018 article in Nature, ICT as a whole contributes around 2% of global emissions, about the same carbon footprint as aviation. (Data centres consume about 1% of all power generated globally, although they contribute only 0.3% of carbon emissions). What’s really worrying, though, is that some models predict that this 2% figure will rise to 20% in the next decade or two.
Let’s be clear here – digital services are more efficient than non-digital in terms of carbon, cost, and time. And making use of cloud computing to deliver those digital services is a good thing: research shows that larger-scale compute and storage has a lower carbon footprint. Consider the huge contribution to energy efficiency from resource sharing: a typical PC user will only use a fraction of the computing power of their device. But individual devices cannot share tasks and will still consume significant energy when idle, thus creating significant greenhouse gas emissions.
By contrast, cloud services allow compute-intensive tasks to be scheduled to start and run at times utilisation is low. Much more importantly, the ability to serve multiple users with far fewer machines reduces the number of machines that need to be manufactured – a significant source of the emissions from IT equipment. When as many as possible use-cases are combined in a resource-sharing data centre, the energy efficiency and avoided fabrication energy and carbon benefits go up dramatically.
But once an organisation has migrated workloads to the cloud, is there more it can do to cut emissions? There certainly is, and what’s more, they can cut their cloud costs at the same time.
Three ways to cut the carbon footprint of your cloud usage
One simple step is to switch server instances off when you are not using them. This is rarely done in private data centres because there aren’t other users queuing up to use each server. But in a public cloud data centre, it frees up that resource for someone else to use – and you stop paying for it.
A slight variation on this is resizing instances according to demand, also known as vertical scaling. If demand for a particular application or service falls, scaling down the server supporting it can free up unneeded resources – although this can be disruptive as it requires the machines to be restarted.
If you have the software development capabilities, rearchitecting applications to be “cloud-native” delivers considerable returns in both cost savings and carbon footprint reduction. By using containers and “serverless” technologies, you maximise resharing of resources, eliminate wasted server time and thus minimise the carbon footprint of your applications.
If you leave everything running ‘on-demand’, you are doing the cloud equivalent of showing up to a restaurant without a booking.
Thirdly, planning your cloud usage in advance can also reduce emissions. If you leave everything running ‘on-demand’, you are doing the cloud equivalent of showing up to a restaurant without a booking. The hyperscale cloud providers never want to turn away a customer so they hold a lot of excess capacity. That capacity can be reduced if customers forecast their future usage and book resources in advance, through reserved instances or savings plans.
There are downsides to this last approach, however. First is the complexity of managing cloud pricing: there are several price points for the myriad of available services, each of which varies by region, term and payment type. Secondly, cloud services are always evolving to stay at the cutting edge, and long-term commitments may prevent organisations from accessing the latest innovations.
To keep track of all the changes, technical or otherwise, that may impact costs and other financial aspects of cloud usage, subscribe to Strategic Blue’s podcast, “What’s New in Cloud FinOps?”