Sustainability Focus: Cloud Efficiency, Not Carbon Emissions
There are many globally pervasive applications today — Netflix, Zoom and TikTok, for example — that use large amounts of computing power. Additionally, the carbon footprint of a number of innovative and important technologies, like IoT, are significant. So yes, carbon emissions of data centers and growing internet traffic is, of course, a problem.
However, it’s not something a developer or someone else from the tech community can directly impact. Where the collective “we” should be focused on is reducing the resource needs of the cloud-based digital products we code today and tomorrow.
This article will take a look at hidden carbon costs, risks associated with the rising energy consumption of data centers, what the major cloud providers are already doing and how resource efficiency in the cloud is achievable.
Is Carbon the Problem?
Conflicting studies confuse the issue of carbon emissions and the data center. One study finds that greenhouse gas emissions (GHG) of data centers within the United States are responsible for only 31.5 million tons or 0.5% of total GHG emissions.
However, another study from Yale finds that the global GHG emissions by data centers amount to over 2% of the human-influenced carbon emissions, equivalent to the global aviation industry, and over 2% of global electricity production.
A third study states that estimated emissions, like the previously cited reports, are not accurate and point to the lack of bottom-up studies enabling precise alignment of the energy source used for data centers worldwide and the carbon emissions of this energy source. Overall, results are inconclusive.
Fortunately, regardless of what the data may or may not be telling us, all the major cloud providers have committed to using renewable energy and invest heavily in much more efficient hyperscale data centers. They have demonstrated prioritizing infrastructure optimization and scale in the ways computing facilities are designed, managed and powered.
Initiatives at AWS, Google and Microsoft have been well documented. While their efforts for cleaner cloud technology are commendable, there remains a problem: Data centers have to keep diesel-powered backup generators for their power plants, and diesel fuel has only a shelf life of 6 to 12 months under the best conditions. Big players like Microsoft are trying to find better solutions to this carbon-intensive backup technology. Natural gas and hydrogen fuel cells are candidates for a cleaner path, but these approaches are still in the pilot phase.
Lack of Renewable Energy Storage Technology
Despite many carbon-related issues with cloud technology, it is not the carbon emission, but rather the rapid increase in energy demand to run cloud-based applications that’s worrisome. As more and more tasks are digitized and much more computing-intensive applications go mainstream, the efficiency gains of Moore’s Law could be outrun by market demand and thereby also outpace renewable energy production. Chip and circuit design can help “flatten the curve” by becoming more efficient, but will run into the limitations of physics soon. How will future data center energy demand evolve as a consequence of its growing use, exacerbated by the fact that there are no storage technologies today at the scale required to store electricity gained from renewables?
To further complicate matters, the digital realm does not exist in isolation. Other energy-intensive use cases like electric cars compete with software on the energy demand side and are already hitting the limits of today’s grid.
The Future Hinges on Resource Efficiency in the Cloud
Everyone would like to contribute to fighting climate change and making it easier for humanity to have a sustainable future. How can the tech community do their part? Amazon chief sustainability architect Adrian Cockcroft has made an important observation about the cloud that cannot be ignored today. Which is, as we’ve successfully used the cloud for digital transformation, why not also use it for sustainability transformation?
To achieve this, the focus needs to be on cloud efficiency. Today, cloud resources needed for running applications are abundant and easy to provision, and are therefore prone to be wasted. Overprovisioning computing resources in the cloud is not a new phenomenon. The FinOps life cycle provides a proven framework for pursuing cloud efficiency.
- Step 1: Inform. First, it’s important to get visibility into your current state to identify the best opportunities for efficiency gains. This entails getting an in-depth understanding of resource allocation and actual utilization at all levels of the compute stack.
- Step 2: Optimize. While the Inform phase shows you the extent of your resource inefficiencies, the Optimize phase tells you what to do about it. This is not always easy given the complexity of today’s digital ecosystem. The good news is that cloud waste can now be identified and remedied using machine learning-based optimization that interprets the signals of your systems and recommends changes for peak optimization around cost and performance.
- Step 3: Operate. Optimization is not a one-time activity. With the dynamic nature of today’s cloud native applications, it requires constant vigilance to ensure peak efficiency. Real transformation requires instilling a culture of sustainable performance and efficiency throughout the teams that are building and operating your applications.
As members of the tech community, we have a responsibility to make sure that applications run as resource efficiently as possible to give us as much time as possible to create the energy ecosystem we need for a sustainable, digital future.