Policy Considerations Towards Converged HPC-AI Platforms

Published On: October 2018Categories: Canadian Science Policy Conference 2018, EditorialsTags:

Author(s):

Monique Crichlow

Director Policy & Strategy Development, Compute Ontario

Chris Loken, Ph.D

Chief Technology Officer, Compute Ontario

Monique Crichlow, Chris Loken, Ph.D

In the 2018 federal budget, the government of Canada committed funds to implement a national Digital Research Infrastructure (DRI) Strategy. Central to the success of the national DRI strategy will be the ability of policy makers to support a system of governance that: a) is responsive to researcher needs, emerging tools and techniques; and b) encourages collaboration with the growing number of stakeholders who make-up the research and innovation sector. The emergence of converged High-Performance Computing-Artificial Intelligence (HPC-AI) platforms as a competitive research tool, presents a tangible opportunity for validating approaches towards a national strategy.

HPC has a long-standing history of supporting advanced research techniques at large scale. In fact, the use of big data on HPC platforms – which is now common place – can be viewed as a first instance of convergence in the digital research era. As more research domains begin to take advantage of big data, HPC scientists have demonstrated responsiveness in their ability to teach and apply advanced computational methods throughout the broader scientific community. With increased interest in, and use of, artificial intelligence and machine learning techniques across research domains, the ability to bring together the benefits of HPC with AI analytics is optimal. “Summit”, the world’s fastest supercomputer, procured by the US Department of Energy, has been positioned as an integral component in achieving the US government’s Artificial Intelligence for American Industry initiative. Its prominence reinforces views from researchers and the vendor community in the ability of converged HPC-AI platforms to scale and accelerate the rate of discovery.

Within the Canadian context, Ontario has already taken actionable steps towards implementing HPC-AI converged platforms. Compute Ontario, a not-for-profit organization, fully funded by the Government of Ontario, works collaboratively with stakeholders to coordinate advanced research computing and big data investments across the province. As a strategic priority endorsed by its partners, Compute Ontario is consciously tackling a known policy obstacle within research and innovation agendas. By focusing on improving access to skills development and big data, and coordinating the distribution of technology investments to leverage the diverse capabilities of the Ontario’s academic HPC consortia, the Province stands to benefit from a more resilient and adaptive DRI ecosystem.

The Health Artificial Intelligence Data Analysis Platform (HAIDAP), a collaborative initiative involving Compute Ontario, the Vector Institute, HPC4Health and the Institute for Clinical and Evaluative Sciences, is a clear example of how the diverse expertise among Ontario’s research community can be harnessed to support emerging needs, while addressing shared priorities. By placing the province’s rich longitudinal health data within a converged HPC-AI environment, the expertise of researchers from across organizations is being coordinated to accelerate the application of AI and machine learning techniques in health. Not only has HAIDAP led to new skill development and greater technical and research capabilities, but also a better understanding of the way in which organizations will need to evolve to meet future demands. Discussions with peers in Quebec (specifically at Calcul Quebec), where similar advanced research assets can be found to those in Ontario, have reinforced HPC-AI convergence as a prime opportunity to accelerate this field. However, it has also emphasized the need for a conducive federal policy environment, that is responsive to support scale in a way that allows all Canadian researchers to benefit from new approaches.

It can be argued that if Canada truly believes a healthy research ecosystem is imperative to remaining competitive in a digital and knowledge-based global economy, then federal policy must also support researchers’ access to robust advanced computing and big data resources. In summary, policies must keep pace with the evolution of research tools and techniques in order to support Canadian innovation and competitiveness.

1. Neely, J.R. (2014). “The U.S Government Role in HPC: Technology, Mission and Policy”. Paper presented at the Conference on Comparing High Performance Computing in the U.S and China, San Diego, CA April 2014. Accessed October 4: https://e-reports-ext.llnl.gov/pdf/773676.pdf
2. Ponce, M et al. (2016). “Scientific Computing, High-Performance Computing and Data Science in Higher Education”. Accessed October 4: https://arxiv.org/pdf/1604.05676v2.pdf
3. Oak Ridge National Laboratory. (2018). “ORNL Launches Summit Supercomputer”. Accessed October 4: https://www.ornl.gov/news/ornl-launches-summit-supercomputer