Enabling Discoveries with AI and HPC (and the Rise of Helium)
This week I am attending the International Supercomputing conference in Frankfurt, and I am in awe of the scientists and researchers that are here and their ability to dig in and understand super complex problems in very specialized areas. While I am humbled by the world-changing work represented at a conference like this, I am also honored to be playing a small part in their success. With the next iteration of SUSE Linux Enterprise High Performance Computing 15 SP1, we’ve expanded and refreshed our bundle of popular HPC tools and libraries that we make available along with every subscription to our SLE HPC operating system.
When most people think about supercomputing, they think hardware. But more and more, there is a shift happening to the software infrastructure and tools needed to support the new wave of AI/ML applications and advanced analytics that are so important in today’s business. Having a strong software platform and tools that enable easier management and monitoring of parallel clusters and HPC workloads are necessary for data-intensive simulations as well as analytics. And just as important is having an HPC software platform that works across heterogeneous hardware environments, from the edge to the core to the cloud, and from IoT to CPUs to GPUs.
When I was in college (yes, that was quite a long time ago) my main astrophysics project was around stellar spectroscopy, where we could infer the physical and chemical properties of stars. Stellar spectra can reveal relative motion, temperature, chemical composition, density and luminosity. What I found most interesting was the variety of celestial objects that could be studied in this way – each with their own story to tell. While we had rather “primitive” supercomputing at the time, HPC provides an environment that enables faster discoveries and helps us realize more applicability here on Earth. For example, the element helium was first discovered in the sun through stellar spectroscopy, before it was found on earth. An unknown line in the sun’s spectrum was observed which could not be related to lines of elements then known in the laboratory. It was named helium after the Greek word for the sun helios, and was subsequently discovered on earth forty years later! That was just an interesting side note – my real point is that discoveries like that rely on having a powerful HPC infrastructure and tools that ease the management and monitoring of those workloads. It all adds up to helping researchers see further than ever before. Today, we rely on helium in cryogenics, superconducting magnets and MRI scanners. It’s a vital noble gas that was only discovered initially by studying solar radiation. Helium is actually rare on earth, produced deep in the earth by radioactive decay. Once it hits our atmosphere, it’s gone forever. But with all of its uses today, we can thank HPC for discovering it and its applicability.
I pick on the second lightest element in the periodic table to make another point. Discoveries at the smallest scale can sometimes lead to the biggest applicability that impacts our everyday lives. The IT infrastructure and tools that we use power a new generation of AI/ML, simulation/modeling and analytics applications. The HPC platform is an important factor for any business or research facility. And, especially in the commercial world, where HPC environments are being built in-house or in the cloud, it becomes even more important to ensure that the open source platform and tools used are vendor-supported to maintain peak performance and utilization. That’s what we strive for with each new iteration of SLE HPC and the bundle of associated tools and computational libraries that we deliver.
Along with our operating system designed for heterogeneous parallel computing environments, we provide tools like slurm for workload management and job scheduling and ganglia for workload performance monitoring. SUSE Linux Enterprise High Performance Computing 15 SP1 continues with that success, adding to the set of popular HPC tools that SUSE supports as well.
Check out my June 11th webinar on AI/HPC in Practice for more info on the AI/HPC market and industry use cases.
And also we have a new white paper available on HPCwire called HPC Meets GPUs: A Love Story.
Thanks for reading my blog post!