Learn about innovative data center storage solutions in this technical talk from industry experts at Supermicro and NVIDIA. Explore a serverless implementation of scale-out All-Flash storage architecture utilizing DPU-based JBOF for storage data management, specifically designed for power-constrained large data centers. Discover how this solution addresses high-performance workloads, including AI training, while examining key implementation challenges such as storage performance optimization, JBOF-DPU integration, data path processing offload support, and multi-protocol network compatibility. Gain insights into EDSFF media form-factors for enhanced density and cooling efficiency, and understand how developers collaborate with the OCP community to validate various storage workloads including parallel file systems and object storage.
High Performance Data Center Storage Using DPUs - Optimizing Power and Performance
Open Compute Project via YouTube
Overview
Syllabus
High Performance Data Center Storage using DPUs
Taught by
Open Compute Project
Related Courses
-
Implement and manage storage for Azure administrators
-
Run high-performance computing (HPC) applications on Azure
-
Next-Generation Storage Using DPUs Instead of CPUs
-
DPU-Accelerated Petascale Storage Subsystem for Near Data Computing
-
Building an Object-based Storage as a Service Solution with Poseidon Storage
-
Cloud Storage Acceleration Layer (CSAL) - Optimizing Cloud Storage Performance with Intel Optane and QLC Flash
