HPC at the core of green and digital transition
Irina Kupiainen, director for public affairs, CSC – IT Center for Science in Finland discusses the role of HPC in driving green and digital transformation.
The dual transition — green and digital — is outlined as one of the cornerstones of post-corona recovery and a source of growth and competitiveness for all industries. The fact that the green and the digital are instrumentally intertwined leads to a new paradigm of thinking about research, innovation, industry, and business. When it comes to the burning need to fight climate change, we must think about not only the absence of negative impact, but about how to turn the demand for climate actions into growth and highlight the positive implications. High Performance Computing is at the heart of this new paradigm, and not only in the technical sense but in various dimensions.
As we know, data is the raw material for digitalisation and an endless source of new innovation and business. It is clear that to reap the benefits of the data expansion, there needs to be efficient enough capacity for processing data, and this is why it is essential to make sustainable long-term investments in HPC. In addition to hardware, we need to invest in software and computing environments, but also people, collaboration, skills development and trust-building — and in R&I ecosystems that enable also new fields of research, such as social sciences and humanities, to make use of HPC resources. Cross-disciplinary data-intensive research will be a game-changer, by creating a helicopter view of what the world looks like, how we can better understand complex phenomena, and how to address global challenges together.
So, to change the game, we need to combine tools, methods, and large datasets from various research fields. Horizontal collaboration is a key requirement not only in the research context but more generally in society, between different actors and sectors, because data will not provide any of its famous added value in silos. This is instrumental when thinking about how to drive the digital transition forward.
And how about the green transition? It is obvious that the climate impact of the ICT sector is significant, and HPC must also be prepared to play its part in the ever more urgent sustainability efforts that the whole world is embarking on. First of all, it must be recognised that the potential of HPC to have a positive impact on climate and the environment is undeniable. It can, for example, support climate research by modellings and digital twins and optimise processes, thus increasing efficiency in the clean energy sector. This potential must be acknowledged and fully leveraged by supporting related RDI efforts and making sure that they can tap into the available computing resources.
At the same time, we must acknowledge the negative impact and systematically seek ways to reduce the carbon footprint of supercomputers. Considering the vast power consumption of HPC systems, it is imperative we promote energy efficiency by e.g., leveraging new technologies. For example, quantum computing may have the potential in the future to accelerate processes in developing e.g., new medicines and vaccines, and thus provide amasing new possibilities for the wellbeing of all people in a very energy-efficient way. However, quantum technologies cannot replace traditional HPC but rather are a new component to HPC, so they are only a part of the solution.
And even if we are able to reduce the power consumption of computing itself, in the world of exponentially growing data, it is clear that solutions based on energy efficiency improvement of the employed technology will not be enough. We cannot stop there. The aim must be full climate neutrality. Powering the supercomputers with renewable energy, using free cooling, and re-use of waste heat are examples of measures that can be taken. For example, the eco-efficiency of the upcoming LUMI supercomputer, placed in Kajaani, Finland, is in a league of its own: LUMI will run with zero or even negative carbon footprint by using 100% renewable energy and by efficient waste heat usage (The waste heat of LUMI can heat up to 20 per cent of the houses of the surrounding city!), while still being one of the fastest HPC systems in the world.
In addition to energy consumption and waste heat, there are other environmental factors to be taken into account when developing HPC systems. The whole lifecycle of the machine must be in scope: Brownfield construction, modularity and scalability, recycling, and re-using the materials used to build the machines. HPC can contribute to the development of a circular economy if we only want to set the criteria in a way that supports sustainability to its full potential.
Another interesting way of applying circular economy is to consider 'recycling' the supercomputer locations. The lifespan of a supercomputer is only five to seven years, after which all the supporting infrastructure built around the machine also becomes redundant. Therefore, it is sustainable to aim at replacing any decommissioned supercomputer with a new one in the same location where the necessary infrastructure and know-how is already in place.
In conclusion, HPC plays a significant role in driving the transition towards economically sustainable digitalisation. In Finland, the hosting country of LUMI, the government recently published a climate and environmental strategy for the ICT sector. This strategy recognises both the heavy footprint but also the potentially huge handprint of the ICT industry. Thinking about what HPC is about takes us back to the basics: climate research is impossible without scientific computing! Addressing grand challenges, such as climate change, requires data-intensive multi-disciplinary global research collaboration. This makes the concept of the environmental handprint even broader than what you could think of from a purely technical perspective.
It is clear, that we need to keep on investing heavily in HPC and data ecosystems, horizontal collaboration and developing skills, and competencies related to data-intensive research.