Modern Storage for Modern HPC and AI Environments

In today's modern landscape, the requirements and challenges of adding storage to an HPC environment have evolved. Networks are 800G and beyond. HPC data sets are shared or sourced across the WAN. Environments are stretched from data centres to the cloud. Traditional spinning disks with hardware RAID controllers presenting block storage to individual servers may not be suitable for your modern HPC environment. Storage demands continue to grow faster than storage capacities - management and archiving is needed. How do you ensure only the right data is using up your valuable storage? Do you implicitly archive all your data using automation, or do you explicitly archive only the data when you choose? This session covers how storage has evolved, with Flash NAND speeds approaching that of system RAM, shared-nothing architectures, Kubernetes-defined storage platforms, archive workflows and more. From NVMe storage in individual servers, on-premises clustered Flash arrays to software-defined storage that spans multiple DCs, we discuss how the landscape has changed and what technologies can solve these modern challenges.