After joining the Intel Select Solutions Program earlier this summer, Penguin Computing™ is announcing its second Intel Select Solution, a reference architecture for converged High Performance Computing (HPC) and Artificial Intelligence (AI). This fully customizable solution enables HPC customers to leverage their HPC infrastructures and expand AI capabilities, resulting in faster time-to-insights for HPC workloads.
Converged HPC and AI Reference Architecture
Traditional HPC workloads, like running large-scale simulations to uncover the correct outcome, are time consuming processes that require dense computing power. In order to speed time-to-insights and curb costs, customers must narrow the number of possible outcomes to arrive at the correct results. Using creative approaches to improve workflows has never been more critical for customers to further innovation and remain competitive.
For HPC customers working with massive and complex data sets in industries like healthcare, academia and federal government, this new Penguin Computing Intel Select Solution adds advanced AI capabilities into the workflow. With the addition of AI, HPC customers are offered a complete solution that significantly reduces the number of simulations needed to make technological inferences and arrive at an answer. Here’s how it works:
Within the AI pipeline, Penguin Computing’s approach includes four components. First, data is curated, organized and indexed so that the data can be trained on. Second, the ingested data enters the training phase, giving the system an opportunity to learn from, categorize, and identify patterns in the data. Third, after the learning phase, is the inferencing phase where users will be able to drastically narrow the number of simulations needed to arrive at the outcome. Finally, customers have the ability to run HPC workloads and applications while starting with fewer iterations than they would if they lacked these AI capabilities. For example, rather than starting with 10,000 possible iterations, the added AI capability enables HPC customers to begin with only the most likely 1,000 iterations to test on.
Customer Benefits and Applications
This solution is ideal for HPC customers and users that need a better way to expand their resources outside of purchasing more HPC clusters. Time to insight is decreased by AI accelerating the entire workflow to run more efficiently. By implementing AI into HPC workloads, customers need less money, time and even in-house expertise, as Penguin Computing will manage the process from start to finish.
Unlike other vendors, Penguin Computing works directly with customers’ existing workflows to customize solutions to fit exactly what they need and is extremely flexible and nimble based on the customer’s existing environment. The AI used is especially unique as customers can only purchase this solution through Penguin Computing, in partnership with Intel Select Solutions. Penguin Computing can plug in the pieces customers don’t have and likewise, support customers who are building a complete environment from scratch.
Intel Select Solutions and Penguin Computing
Through Penguin Computing’s collaboration with Intel, this reference architecture can be applied across a wide range of HPC and AI applications from federal, to enterprise, to academia. These reference designs are optimized for a range of broadly used HPC and AI/ML applications and every Intel Select Solution from Penguin Computing is built to meet the highest thresholds of resiliency, system agility and service reliability.
The solution will be available for purchase in October, 2020.