Sustainable Big Data Conversations: Tackling the hidden costs of computing

Blog post by Federica Lucivero and Abram Schonfeldt 

On 17th July 2023, the Big Data Institute and Ethox hosted the inaugural event of the Sustainable Big Data Conversations.

Conceived as an output of the project Digital Sustainability, this series aims to engage the Big Data Institute (BDI) and Oxford Population Health (OxPop) research community with the topic of the environmental sustainability of their research. Dr Loïc Lannelongue and Prof Michael Inouye (University of Cambridge) who have performed pioneering research and are key players in the field of green computing joined BDI to deliver a talk titled “Tackling the hidden costs of computing: GREENER principles for sustainable computational science”.

Prof Inouye explained how, as wildfires were burning land and homes and killing animals and people in Australia in 2019/20, he felt the urge to make climate change a central issue across sectors and asked the help of his interdisciplinary team in his system genomics Lab. Loïc, a PhD student at the time, was one of the volunteers who responded to this call for action. This is how the Green Algorithms project started.

As Loïc eloquently explained in his talk, high-performance computing offers great opportunities to improve human health and the environment, but it is not free. It comes at an environmental cost.

Three types of environmental impacts should be considered: the lifecycle footprint of the hardware, the long term data storage and the powering of computers.

The Green Algorithms tool quantifies the carbon footprint of computation and can be used to raise awareness and facilitate greener computation. Monitoring environmental impacts can be a first step to minimise the energy needs of computation, but it is not enough: it has to be combined with synergic action involving different stakeholders, embracing their individual and institutional responsibilities, engaging in collaborations to facilitate equitable access to low-carbon computation, designing and delivering training programmes and conducting research on this topic. These GREENER principles can guide actions for sustainable computational science and contribute to a cultural change that makes environmental sustainability a core element of research.

As Loïc concluded, “there are many low hanging fruits in this field, which means that there is a lot of space for action. Simple actions, such as using a laptop until the end of its life instead of replacing it, or debugging software on small datasets, can have big impacts if taken up at institutional level.”

At the end of the talk, Prof Cecilia Lindgren, BDI Director, reflected on the need for an institution like the BDI, where most groups engage in intense computations with the goal of improving health globally, to raise awareness and act to reduce the environmental impact of the research conducted. She addressed the need to do so at different levels, ranging from changes that individuals can make to their research practices to broader practices adopted at the group/institute level up to actions at the level of funding bodies and journals.

The presence of a diverse group of people, including early career researchers, PIs and scientific computing teams, facilitated a discussion of the actions that we can take at different levels. At the individual level, researchers could use their personal computers as long as possible, use freely available tools to calculate the carbon usage of common analyses, report their carbon usage on the cluster to their PIs, doing small “pilot runs” of analyses before running them at scale, and reporting an estimation of the carbon impact of the results obtained in paper submissions.

Hopefully, this awareness at the individual level will provoke, and subsequently align with, awareness at institutional and higher levels that the impact of research needs to include its impact on the climate. Many of these changes require the setting of priorities, the set up governance mechanisms (or the revision of existing ones), and to address issues of fair and equitable research practices. These are qualitative questions that can be addressed through an empirical ethics approach.

The lively conversation and high attendance (people had to join online because the room was at its maximum capacity) demonstrate a high interest within the BDI/OxPop community to these address these issues. This event was part of a series of conversations that will bring key stakeholders (funders, University central offices, lab certification providers, ethics committee’s representatives) to engage in conversations with the BDI/OxPop community on issues of environmental sustainability of their research.

If you participated in the event, we would like to hear any action points that you brought with you walking out of the room.

To be informed about future events in the series please visit the Sustainable Big Data Conversations web page or email Dr Federica Lucivero.

Links to meeting resources:

• Slide presentation, Dr Loïc Lannelongue and Prof Michael Inouye
• Green Algorithms project (where you can calculate the carbon footprint of your computations)
• GREENER principles article