Fabric Attached Memory: Hardware and Software Architecture for HPC Systems

Watch a 19-minute conference presentation from the SNIA Compute+Memory+Storage Summit exploring fabric-attached memory (FAM) architecture for large-scale data processing in HPC environments. Discover how emerging AI and data analytics workloads are driving innovations in hardware and software architectures, with a focus on CXL technology and high-performance data analytics. Learn about a novel system architecture that enables large datasets to be accessed across Slingshot-connected HPC clusters through fabric-attached memory, bridging the gap between traditional HPC, AI, and data analytics applications. Follow along as HPE's Clarete Riana Crasta demonstrates the practical implementation through an Arkouda-based proxy application for real-time analytics, covering the OpenFAM software stack, component architecture, access latency considerations, and future development directions.