The complex layers of cloud computing sustainability

I’ve argued over the years that cloud computing is a step in the right direction when it comes to sustainable computing. My viewpoint often opposes environmental organizations that argue against the many new power-hungry data centers that cloud companies build.

The sustainability of public cloud computing is easy to understand. Simply put, cloud does more processing and storage with the same number of physical servers and data centers. How? Cloud computing’s multitenant approach puts more applications, data sets, and users on a smaller amount of hardware, at approximately 85% to 95% utilization of capacity. Compare this to traditional approaches where we own the servers and data centers, and the hardware resources are often utilized at a very low capacity, often 3% to 7%.

Cloud requires less power for the same amount of processing. Does this always make cloud computing green—or should we say, “greener”?

The word “always” rarely turns out to be accurate. It’s the same with cloud—the issues are rarely black and white. To investigate cloud sustainability, let’s break it down into the two major layers.

Source of power

People often brag about how their electric car has a zero-carbon footprint. It’s not that simple. Most power, at least in the USA, is generated by burning fossil fuels. In 2021, it was 60% fossil fuel with the remainder split between nuclear and renewables. No matter where you charge your Tesla or power your cloud or non-cloud data center, power consumption has a carbon impact.

The argument for cloud computing sustainability is that it reduces the required hardware and data center space. However, both cloud and non-cloud application processing requires power grids driven by fossil fuels. 

The opportunities to reduce cloud-related carbon not only depend on the use of sharable IT resources in public clouds but also the locations of the data centers for public clouds. The trend is to have points of presence as close as possible to those using the cloud services. Many of those points depend on carbon-heavy power sources. In those instances, cloud computing is not as green.

You could make the argument that it’s better to leverage an enterprise-owned data center with much lower utilization of server resources that’s in an area served by renewable and/or nuclear energy. In that instance, it’s not as green to choose a cloud computing provider’s point of presence that only uses carbon-heavy generating power.

Optimizing cloud solutions and resources

No matter if your power burns fossil fuel or not, much can be said about how you optimize the cloud resources you use. For example, if you give two developers the same business problem to solve using public cloud resources, you’ll usually find that one of them does a much better job at leveraging the minimum number of resources for the maximum effect in terms of value returned to the business. 

This can have very different impacts on resources used to solve basically the same problem. For example, the fully optimized version uses only three compute resources and two storage resources. The second uses three times more to solve the same problems. Thus, the second solution also burns approximately three times the power. 

Remember: There are cost and carbon impact penalties if you put underoptimized solutions on a public cloud. The more unnecessary resources, the more unnecessary cloud fees. Depending on how your public cloud provider powers its data center, this could also have a significant impact on carbon output. 

It’s complicated

You can do audits yourself or use an outside organization to dig into the specifics of how your use of public clouds helps or hurts the planet. These audits include an explanation of the ultimate source and location of the power as well as how you specifically use the cloud resources.

The findings often surprise enterprises. Some believe their use of cloud computing is at the height of sustainability. It could turn out that their cloud solutions are highly inefficient if they just lifted and shifted applications and did not refactor them for cloud resource optimization. Or an enterprise might not understand the true source of the power. It could be as simple as just moving their workloads to another region that only uses renewables as a source. Yes, there might be some impact on latency, but the positive impact on sustainability might offset it. 

This is interesting stuff. I agree with using cloud computing to reduce carbon. However, like anything else, we need to understand what’s truly happening before we declare success.