OU blog

Personal Blogs

Christopher Douce

Writer greener software

Visible to anyone in the world

I spent a bit of the morning of 2 April 25 listening to a presentation by Charles Humble on the topic of Writing Greener Software (YouTube).

It is a good listen watch, especially if you’re interested in software. To pique your interest, the following topics are mentioned:

  • Some science behind the study of climate change (along with some terrifying predictions).
  • Different ‘scopes’ of carbon responsibility: scope 1 (what we are responsible for), scope 2 (carbon from some of the services we consume), and scope 3 (carbon emitted from organisations we use). I paraphrase this badly.
  • The notion of embedded carbon and how it relates to your devices.
  • Related to this, keep your end user devices for as long as you can, since the embedded carbon is likely to be higher than any running costs.
  • The relative efficiency of programming languages. With this in mind, always try to use the latest runtime of any language you’re using (a point was made that new version of the Java virtual machine is likely to be more energy efficient than older version).
  • The efficiency of servers (and what server load you should aim for). Also, when it comes to energy efficiency, servers are different to end user devices.
  • The need to manage our ‘zombie’ servers; servers which we’re afraid to turn off. On this point, we need to be able to reliably turn servers off and on without causing any disruption to services.
  • The CPU choice of your data centre servers.
  • If you’re using a public cloud provider, to look at any carbon tools they have, and know how to use them.
  • There’s a concept called demand shifting. Consider where your users are, and consider where the energy is greener. You could also carry out your data processing at different times, and at different locations. Try to predict demand.
  • On the topic of AI, ‘information on the environmental cost of AI is scant’. Energy consumption may have increased between 30 and 50% since 2020. Generative AI is computationally expensive, both with respect to training and using large language models.
  • There are likely to be more regulations and governance.
  • Conclusion: share your knowledge about all this; talk with each other.

Accompanying references

I recognised a couple of websites and resources, and learnt about a few that I hadn’t heard of.

In terms of websites, tools and organisations, there were:

Some useful books:

There are also a couple of podcasts (but I can only access one of them): Climate Stack.

Reflections

I was expecting something different from this talk. I was expecting to learn some practical tips that perhaps related to design and coding. In retrospect, I’m really glad it didn’t touch on these lower level issues, since that wouldn’t express the complexities that the talk touches on. Green software goes beyond bits of code. Software relates to requirements, architectural choices, data centre locations, data centre design, and decisions about when and where you carry out tasks. There’s also the importance of personal and professional responsibility, and the need to lobby within the organisations in which we work.

Permalink Add your comment
Share post

This blog might contain posts that are only visible to logged-in users, or where only logged-in users can comment. If you have an account on the system, please log in for full access.

Total visits to this blog: 2622604