The first season of HPC Thursdays webinar series comes to an end. While waiting for the second season to kick off, here there are the main take-aways from these last months.
High-performance computing (HPC), also known as supercomputing, is an innovation accelerator that drives significant financial returns for companies and enables a leap forward to research as well. Luxembourg has a unique HPC infrastructure that is open to companies as well as to researchers.
Even though high-Performance computing is common in the research domain, HPC is still very little known in the private sector. To fill this gap, the HPC competence centre in Luxembourg organised a series of webinars, targeting newcomers in the supercomputing world, to provide an insight into why HPC is relevant and useful for businesses and research, and how it can be used.
The webinars provided concrete examples of use cases in different sectors and introduced the experts who can help you get your own projects off the ground. Let’s see the main highlights of each of the 8 webinars.
The benefits and impact of HPC
Although there is not one single metric that can be used across all sectors to measures the benefits of HPC, it is clear that it can benefit companies of all types and sizes in all sectors. The reasons for using supercomputing tend to be individual for each organisation and related to its strategic priorities. The best metric to justify HPC is “time to solution” and “inability to solve the problem by any other means”, among other valid reasons.
While HPC is already today driving business and research competitiveness, its importance will increase exponentially in the future as the amount of data available doubles every two years. There is a great potential for incorporating the exploitation of data into the business models of many companies. A study by Hyperion Research published in November 2020 analyses 763 HPC projects from academia, government and industrial sectors. The study finds that high performance computing generates $47 in profit/cost savings for every dollar of investment in HPC systems. The Luxembourg government heavily invested into HPC infrastructure in order to enable businesses and research to flourish by using these cutting-edge technologies.
Having some basic knowledge of how high performance computers work obviously makes it easier to plan and prepare projects. This webinar covered everything from CPUs (processors of roughly the same family as those that can be found in an ordinary laptop or PC) to accelerators such as graphical cards (also called GPUs) and a more exotic card type (FPGA), and memory capacities.
The speaker shared some basic information about software packages available on supercomputers, and about how to choose the appropriate programming language or programming model. However, a key starting point for successful HPC projects is a clear understanding of the problem that is to be addressed (in the fields of physics, biology, finance, mechanics, data analysis, etc.) as well as a rough estimation of the needs in terms of computing resources. A good first step is running some trial-and-error jobs on an ordinary computer as a way to estimate how many nodes and how much memory, wall time (the time needed to run an application), etc. would be needed.
This webinar focused on business and research benefits of supercomputing simulation at scale, with LuxProvide’s expert Alban Rousset, HPC senior solutions engineer, as the main speaker. Digital simulations are used in many fields, such as biology, seismic processing, climate and rendering.
There are several reasons why HPC-enabled at-scale simulations are useful:
- They quickly produce results and improve the time to market or solution
- They are able to tackle larger problems by moving at a larger scale and increasing the accuracy, granularity and resolution
- They help save costs in terms of money, time and manpower
- They solve problems that just cannot be solved properly or correctly without using a supercomputer
The speaker also showcased concrete results obtained with supercomputers, and in particular with MeluXina. The examples ranged from industry (e.g. digital twins) to sustainability (carbon emission reductions) and the simulation of a flock of birds, employed in biological transformation (i.e. the application of the materials, structures and principles of living nature to technology and management in order to achieve sustainable value creation).
Data analytics and artificial intelligence on a supercomputer
The 4th webinar explored data and artificial intelligence (AI) with LuxProvide’s expert Luis Vela Vela, HPC senior solutions engineer. He started from the basics: data, a resource that is growing exponentially in volume. “If we’re not already drowning in an abundance of data, then we certainly will be in the future,” he said. Dr Vela then explained the meaning of AI, the relationships between the data and the vast palette of AI techniques that exist.
He also provided examples of AI applied to business with three concrete use cases:
- predictive maintenance
- natural language models
- reduced order models
HPC for health – journey experience testimonial
For this webinar, we invited PhD student Eleftheria Charalambous who shared her HPC journey in the biomedicine domain: how it started, the challenges she faced, the expectations she had, and the results she obtained.
She clearly highlighted why using HPC was crucial for the success of her research and how much more she will need HPC in the future for achieving even more. Finally, as a source of inspiration, she provided examples of how HPC can be used in the health field, highlighting applications she sees in the health sector beyond just biomedicine.
HPC for sustainable finance
Emmanuel Kieffer, research scientist at the University of Luxembourg, presented artificial intelligence (AI) in action in fintech, and in particular how to use AI to recommit capital in private equity. Capital committed to private equity funds is irrevocable and expose investors to cash flow uncertainty and illiquidity. Recommitment strategies are still manually designed, which is a tedious and critical task.
Dr Kieffer illustrated how HPC can provide new opportunities in this field too: use reinforcement learning to discover reliable capital commitment policies (i.e. size and timing of capital calling) so that optimising returns and maintaining the targets of limited partner investors.
HPC for digital history and public administration
Two HPC use case examples were presented. The first example came from the social sciences and humanities, and highlighted the transversal application of HPC to all domains, not only to the usual “tech” disciplines. Lars Wieneke, head of digital research infrastructure at the Luxembourg Centre for Contemporary and Digital History of the University of Luxembourg, illustrated why digital history needs HPC. He explained that there are more and more different types of sources, and very often the value added comes from cross-connecting them and establishing relationships between different datasets. He also pointed out that supercomputing is necessary for building AI algorithms that are capable of creating a reconstruction of old sources difficult to analyse otherwise.
The second example came from Austria and illustrated HPC applied to the public administration. Allan Hanbury, Professor for Data Intelligence and Head of the E-Commerce Research Unit of the Technische Universität Wien, presented the BRISE project, whose aim is to fully digitalise the building permit process from application to approval, thereby reducing the time required to issue a permit by 50%. He presented the crucial role of HPC for training deep neural networks that allow for automated checking of the submitted building plans against building rules in law, so to facilitate the work of building inspectors.
Discover MeluXina, Luxembourg’s unique business oriented supercomputer
The last session was dedicated to MeluXina supercomputer and LuxProvide’s services. LuxProvide is the company that operates MeluXina supercomputer and provides the necessary expertise to support companies to make the best use of it.
Valentin Plugaru, Chief Technology Officer at LuxProvide, presented MeluXina’s architecture and technologies, software ecosystem and platform service. He also gave an overview about the HPC projects that have been run or are currently running on MeluXina.
MeluXina is designed as a modular system (different types of processors, such as CPUs, GPUs, FPGAs, memory) to offer world-class HPC, high performance data analytics (HPDA) and AI services for a wide variety of workloads and application domains. This supercomputer is also unique for its secure environment specifically designed to ensure the security measures required by companies, which want their data to remain private.
All the replays are also available on the Luxinovation’s web site page dedicated to HPC Thursdays.
Pre-register to the 2nd edition of HPC Thursdays webinar series to be informed when it will start again! If you want to see HPC applications in your sector, let us know!