Sustainability of AI-scale digital research infrastructure
Steve Quenette1, Paul Coddington2, Mark Gray3, Angus Macoustra4, Tim Pugh5, Carmel Walsh, Max Wilkinson2, Allan Williams6
1Innate Innovation, Melbourne, Victoria, Australia
2Australian Research Data Commons (ARDC) Australia
3Pawsey Supercomputing Research Centre, Perth, Western Australia, Australia
4CSIRO Australia5Bureau of Meteorology (BOM), Melbourne, Victoria, Australia
6National Computational Infrastructure (NCI), Canberra, Australian Capital Territory, Australia
Abstract
A decade of the cloud movement has inadvertently produced global year-on-year data centre energy efficiencies. Over this period, total data centre electricity use has barely grown, but we are using over 500x more (virtualised) servers. Recently the world has been captured by a disruption popularised by ChatGPTv3 – how Large Language Model forms of deep learning AI will imminently transform business, society and research. Unlike the Web-2.0 era, adapting LLMs to an organisation or project is compute-intensive, requiring dedicated use of traditionally HPC-scale facilities. The electricity needed for data centres will explode, raising significant environmental and cost questions. Are we ready for this?
This workshop covers three distinct groups of our ecosystem: large research-centric datacentres, exposure to emerging sustainability (ICT) technology vendors and research organisations. It will educate on the AI-driven transition and how this will affect the digital infrastructure landscape and inform sovereign direction. It will showcase some emerging approaches & technologies to mitigate and alleviate. It will be interactive in assessing the current awareness and situation. A summary report will be openly published.
The official page for the workshop is https://www.innateinnovation.co/workshop23, where you can learn to expect, who is participating and other further developments.
Biography
Steve Quenette is the founder and principal of Innate Innovation, which provides technology & data governance advice to socially responsible, innovative organisations. He is the former Deputy Director of the Monash eResearch Centre at Monash University. He also contributed to national digital research infrastructure by founding the Australian Research Container Orchestration Service (ARCOS), the Australian Scaleable Drone Cloud (ASDC), and the Research Data Cultures Conversation (RDCC). He established and operated a Nectar Research Cloud federation node for ten years.
Steve’s areas of expertise include business strategy, technology strategy, data governance and operations.
Paul Coddington is an Associate Director at the Australian Research Data Commons (ARDC), leading the national Nectar Research Cloud. He has more than 35 years experience in eResearch including cloud computing, high-performance and distributed computing, computational science, software development, data repositories, research platforms, and project and program management.
Mark Gray is currently the Head of Scientific Platforms at the Pawsey Supercomputing Centre. He brings a strong research and data management background to his role of building the next generation of supercomputing systems for Australia. This includes procurement, deployment, training, user expectation management – and leadership of the team who administer and operate Pawsey systems.
Mark has a background in earth science and research software engineering with experience at the University of Wisconsin-Madison, NASA Goddard Space Flight Center. Prior to joining Pawsey, Mark worked with the Terrestrial Ecosystem Research Network and the Integrated Marine Observing system in Australia.
Tim F Pugh has 34 years of experience in the field of scientific and high-performance computing. Today, his managing role is in HPC Planning & Capability at the Australian Bureau of Meteorology. Tim has accumulated 18 years of experience at the Bureau of Meteorology in investment planning and program/project management to software engineering and design of scientific applications for research and operations.
Carmel Walsh has worked in the Tertiary Sector for over 25 years and has led digital transformation in national and institutional research learning & teaching and leading enterprise IT across government and corporate organisations for the last 15 years. She most recently led the national strategy and service delivery of the Nectar Research Cloud and long-term research data storage at the ARDC.
Max Wilkinson has a comprehensive background in research data management, research data governance and research infrastructure operations. For the last 3 years he has worked with the Australian Research Data Commons as a research data infrastructure architect, designing a scalable and sustainable investment model for nationally significant research data collections. Prior to this, he has worked with the National eScience Infrastructure (NeSI), Council of New Zealand Research Librarians (CONZUL), AgResearch, eResearch2020 and NZ Ministry of Business Innovation and Employment. He lived and worked in the UK for two decades, most recently as Head Of Research Data and Network Services at University College London, the Datasets Programme Manager at the British Library and Informatics coordinator at Cancer Research UK. He received his PhD in Molecular Nephrology from UCL in 2003. ORCID: 0000-0001-7899-1842