Green Computing at RJJ Software: What We Changed

The cover image for this post is by Brian Garrity

Introduction

Green computing, or sustainable software development, focuses on minimising energy consumption and waste throughout a software product’s entire lifecycle. This post covers the measures we have taken, and the resources we drew on:

Energy-efficient Hardware

When this case study was written, colleagues worked on M2 series Mac hardware, which draws less power and generates less heat than the alternatives. Andrew Cunningham measured the difference in his review of the M2 Pro Mini:

As measured by macOS’ built-in powermetrics command-line tool, the M2 Pro’s average power usage when fully loaded is around 36 W, whereas the Core i5 can use between 65 and 150 W and the Ryzen 7 between 90 and 136 W.
- Andrew Cunningham, M2 Pro Mac mini review: Apple’s Goldilocks desktop for semi-professionals

For every Core i5 system running at the higher end of its average power, you could run four M2 Pro Minis within their average range.

Some colleagues have since moved to M4 Pro based machines. The M2 hardware has not been retired, and our reasoning is that we should work off its embodied carbon first rather than replace machines that still do the job. Apple’s own environmental report for the MacBook Pro 14-inch with M4 Pro puts production at 74% of its 218 kg lifecycle footprint, against 20% for the four years of use it assumes. Those emissions are spent once the machine exists, so we would rather keep it in service than trade it for a more efficient model and start that cost again.

Optimised Code

Efficient code matters in sustainable software development. Our developers focus on creating efficient algorithms and eliminating unnecessary computations, which reduces the overall resource usage of their applications.

Code optimisation varies with the application and the codebase. It helps to start with a “heat map” of the most frequently used features, so you know which areas are worth the effort. Measuring the codebase against hardware representative of the live server, and against real user actions, gives a more honest picture than a developer machine will. It is also worth auditing the third-party libraries in use and looking for alternatives with better performance.

We would recommend taking a look at the Green Web Engineering research paper by Markus Dick, Stefan Naumann and Alexandra Held for optimisation suggestions for web-specific applications.

Decentralisation with Tailscale

Adopting a decentralised architecture, such as using Tailscale for a VPN solution, allows us to avoid large centralised data centres that consume high amounts of energy. This architecture is coupled with our remote working policy, described below. It lowers our total energy consumption as a company, and it improves privacy and security for colleagues.

Tailscale sets up a peer-to-peer VPN across all our devices, so there is no central VPN server to keep running. Devices appear on the secure network only when they are powered on and in use.

Cloud Providers with Green Initiatives

In pursuit of sustainability, we actively investigate cloud providers that prioritise environmentally responsible practices and implement green initiatives. By choosing these partners, we contribute to a more sustainable software ecosystem.

Green Web Engineering, published in 2010, gives a sense of the scale even then:

The estimated power consumption of data centres in the U.S. was increasing from 28 in 2000 to 61 billion kw/h in the year 2006 (U.S. EPA, 2007) and in the world from 58 billion kw/h in 2000 to 123 billion kw/h in 2005 (Koomey, 2007). Consequently, considering problems like climate change, reducing energy consumption of the Web is necessary.
- Markus Dick, Stefan Naumann and Alexandra Held, Green Web Engineering

Two things are worth drawing out of that. The 123 billion kWh is the world figure for 2005; the United States on its own accounted for 61 billion kWh in 2006. And both numbers are now two decades old.

A billion kilowatt hours is a terawatt hour, so they can be set directly against current ones. The International Energy Agency puts global data centre electricity consumption at 485 TWh for 2025, roughly four times the 2005 world figure. It expects that to roughly double again by 2030, to around 950 TWh, or about 3% of global electricity demand (Key Questions on Energy and AI, IEA, 2026).

Remote Work and Reduced Commuting

Allowing colleagues to work remotely from anywhere cuts daily commuting and the carbon footprint that goes with it. It also keeps people off crowded public transport and busy roads.

Mark, then our Chief Podcast Mastering Consultant, emphasised the benefits of remote working in terms of flexibility and convenience:

Being able to work remotely means that I can literally work from wherever I want. If that means that I work out of my home office, the local coffee shop, or a rented office space, I’m able to get the work done without any issues. As long as I connect to the VPN, I can access all of my files and contact whoever I need using our internal chat app.
- Mark Taylor, then Chief Podcast Mastering Consultant

Remote working suits our colleagues, and it takes the commute out of their carbon footprint entirely. For scale, cars and taxis accounted for around 57 million tonnes of CO2 equivalent in 2021, according to the UK government’s Transport and environment statistics: 2023.

Conclusion

That covers the measures we have taken so far: hardware chosen for its energy efficiency, attention to how our code performs, a decentralised VPN in place of an always-on server, cloud providers picked partly on their green initiatives, and a remote-first policy that removes the commute entirely.

You can read more about our green computing initiatives and goals in Our Green Initiatives.