HPC Yearbook 19/20

Road to AI

Rob Johnson highlights the steps you should consider towards AI, and how your existing HPC architecture can help to get there


High-Performance Computing 2019-20

esearch institutions, government agencies and enterprises seek more profound insight from the massive datasets available to them.

According to IDC, over half of the world’s data was created in the last two years, yet less than 2 per cent has been analysed1. For many years now, the melding of HPC and high performance data analytics (HPDA) led organisations to new approaches to extract more meaning from data. Today, though, AI advances and extends, and accelerated these proven approaches. For instance HPDA and AI help companies

design and bring products to market faster. Te combination also empowers faster scientific breakthroughs, helps financial institutions detect fraud, enables self-driving cars and much more. As HPC systems and the technologies underlying them offer new ways to accelerate workloads, firms increasingly embrace AI for its multitude of benefits.

Workload convergence: one system, many workloads

Rob Johnson

In the past, a single workload – such as modelling and simulation – occupied the vast majority of an HPC system’s resources. However, technical advancements following the pace of Moore’s Law make it possible to extract more from each node in a cluster. Modern computing technologies not only process data faster, but they can also minimise data movement for reduced latency and much quicker time to insight. Tese advancements give HPC systems the prowess to run multiple workloads on a single system. A single cluster can also accommodate

converged workloads, combining AI with more traditional workflows like simulation and modelling. Te joint effort can coax more insight than possible with a legacy ‘single- workload’ approach. Since AI can be trained by researchers to look for items of interest using scientific hypotheses, it can help identify patterns and insights impossible for scientists. In the pharmaceutical industry, staff seek

to accelerate the development of new and effective medicines. In the past, the process of identifying beneficial drug molecules proved very costly and time-consuming. However, a trained neural network can evaluate millions of molecular candidates for potential medical application and find those few which offer the most significant promise. From there, scientists can analyse the shortlist for viability. Each company in its respective industry

segment needs to evaluate if, and how, HPC and AI can advance their organisation. Te first step in the journey involves aligning business goals with compute needs. Exploring


questions like how AI could improve a legacy product design process, or how it could make products safer and more energy efficient, and help an organisation make informed decisions about the business opportunity that AI offers.

Data-centric culture

Planning, procuring, implementing and managing a HPC system takes time. An organisation must embrace a ‘data-centric’ culture. C-level executives, IT managers, and other stakeholders must all agree on the importance of embracing – and investing in – new approaches to advance their company. Making effective use of available

information increases business intelligence. In turn, that means faster corporate insights and decisions. One parts manufacturer, for instance, combines Internet of Tings (IoT) devices and AI to monitor machining equipment. When the system notices a slowdown in the spindle’s turning rate, scheduled maintenance can avoid costly, unexpected downtime. Other manufacturers use AI and IoT to monitor delivery trucks to predict potential impacts to the supply chain.

Getting the right team in place

Te next step toward making HPC and AI deployment successful involves gathering the right expertise. While enterprises and research institutions oſten have the benefit of in-house knowledge, others do not. HPC deployment involves experts like

data scientists, developers, line-of-business managers, and others. Organisations should consider if their existing IT team requires additional staff, or if external providers can fill vacant skill-gaps. OEMs and systems integrators working to

address this expertise gap provide simplified solutions to facilitate execution of demanding workloads in an HPC environment. One approach they take involves the use of Intel Select Solutions for HPC. Tese solutions feature pre-configured, optimised and validated systems that provide a more straightforward path to HPC for specialised workloads like simulation and modelling, genomics and more. To meet modern demands Intel also offers a ‘select solution’ for HPC and AI converged clusters. For those institutions just beginning

the journey to HPC, cloud-hosted HPC solutions offer a stellar starting point. Hosted environments offer end-users a fast and turnkey solution to run their workloads. Also, the flexible nature of cloud solutions provides a pay-as-you-go model that allows customers to scale their workloads up or

Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32