With the increased focus of companies on environmental, social and governance criteria, sustainability has become a trending topic. Energy used to power IT workloads can contribute significantly to a company’s environmental footprint, making it important to think about how the sustainability of IT architecture and strategy could be improved.  

In this blogpost we discuss ideas on how to get started with sustainable IT. We also elaborate on guidelines we follow at ML6 to make sure that the sustainable IT mindset is shared across the entire team, from advisory to development.  

‍

The quick fix

Data center energy efficiency is measured in Power usage effectiveness (PUE). Google Cloud regularly publishes PUE information. Azure and AWS are less open about this according to this article by Sustainable Finance. The average PUE is 1.67.

From a high-level perspective, the quick fix for sustainable IT is comparing the PUE of your current data center provider with the PUE of other (cloud) providers and moving certain workloads.
With a PUE of 1.1 Google is currently the leader.

‍

Google is fully carbon neutral and recently started publishing more detailed KPIs by location.

Google CFE%: This is the average percentage of carbon-free energy consumed in a particular location on an hourly basis while taking into account the investments made in renewable energy in that location.

Grid carbon intensity (gCO2eq/kWh): This metric indicates the average lifecycle gross emissions per unit of energy from the grid.

By using the recently introduced Google Cloud Region Picker you can identify the most carbon-friendly region.
For Belgium, it’s interesting that Switzerland and Finland are slightly better than the Belgian location.

‍

‍

‍

Sustainable IT architecture & strategy

‍

Reducing your carbon footprint by lift & shifting a number of workloads is a good start but a lot more can be done.
Sustainable IT has to be a mindset shared with the entire team.

‍

‍

We are currently working on a number of guidelines for our advisory and development team.

‍

‍

1. Design thinking 
   ‍Are you sure you are solving the right problem? We often notice in our advisory services and projects that compute- or data-intensive features and even entire tech stacks are not used a lot a few months after they were introduced. This risk can be mitigated by basic usage monitoring, regularly reviewing your cloud bills and the relevance of your tech stack.
   ‍
2. Scale to zero 
   ‍Aim for infrastructure that can scale to zero compute. On-premise this is difficult to achieve but on a cloud platform, multiple options are available. Ranging from functions as a service, managed container platforms for more complex workloads, cloud data warehouse and SQL on top of distributed storage services that are billed by the second or byte processed. Large data and ML pipelines that used to be run on large Hadoop cluster can be refactored with minimal changes in separate jobs on a temporary cluster. In most scenarios, an easy quick win is investing in development and test environments that can be switch off outside working hours or in case no tests are running.
   ‍
3. Right-size your infrastructure 
   Depending on the phase of a project, the footprint of the infrastructure should be regularly reviewed. It’s not a problem to temporary develop on a more constrained or powerful system. As soon as the real-life workload is known, the infrastructure can be right-sized and monitored.
   ‍
4. Responsible AI & Advanced Analytics 
   Check out our article about sustainable AI. We discuss multiple options to develop, optimize and deploy ML model with sustainability in mind.

‍

‍

Summary

We hope we gave you a few ideas to get started with sustainable IT.
Are you approaching sustainable IT in a different way?
We are always happy to learn about different approaches.
Get in touch if you need support with your cloud/data or AI strategy.

‍