Explore a 42-minute conference talk from SNIA SDC 2022 detailing Los Alamos National Laboratory's groundbreaking approach to scientific data management using NVME Computational Storage. Discover how the laboratory is transitioning from traditional Petabyte-sized file management to an innovative record and column-based methodology. Follow Gary Grider's presentation on three proof-of-concept applications and industry partnerships that are reshaping large-scale data management. Delve into technical aspects including HPC Scientific Simulation Systems, ZFS Checksums, Erasure, server memory bandwidth challenges, File System Services Offload, and the Accelerated Box of Flash project. Learn about specialized applications like Vector Particle in Cell (VPIC), KV-CSA Key Value Computational Storage Array, and the intricacies of Grid Method and Adaptive Mesh Refinement (AMR) for indexing multi-dimensional unstructured adaptive meshes.
HPC Scientific Simulation and Computational Storage for Large-Scale Data Management
Overview
Syllabus
Intro
Background: HPC Scientific Simulation Systems Trinity-circa 2016
Why Offload? ZFS Checksums, Erasure, Compressive Server memory bandwidth is problematic and expensive
How to Consume: File System Services Offload
Accelerated Box of Flash: Powerful Computational Storage for Big Data Projects
Vector Particle in Cell (VPIC) our Record Based/Single Dimensional Index Application
KV-CSA Key Value Computational Storage Array
What's a Grid Method and an Adaptive Mesh Refinement (AMR)?
Indexing Multi-Dimensional Unstructured Adaptive Meshes
Taught by
SNIAVideo
