Solidigm CSAL Solution Brings Advanced IO Shaping, Caching and Data Placement Into NVIDIA DPU DOCA

Explore how the Cloud Storage Acceleration Layer (CSAL) integrates with NVIDIA DPU DOCA to deliver advanced IO shaping, caching, and data placement capabilities for BigData and AI workloads in this 16-minute conference talk. Learn about CSAL as an open-source user mode Flash Translation Layer (FTL), cache, and IO trace component within SPDK that has been commercially deployed to enhance Alibaba's cloud storage system. Discover how this joint development between Solidigm and NVIDIA leverages DPU DRAM as CSAL write buffer to achieve optimal storage latency while maintaining data consistency. Understand the benefits of combining QLC high-density storage with DPU storage solutions for AI data centers, including power and space savings. Examine the integration of advanced storage IO shaping, caching, and data placement software into NVIDIA DPU DOCA storage services, and review experimental data from collaborative work with BeeGFS. Gain insights into this technology that has been recognized in academic research, including joint publications at top-tier conferences like EuroSys 2024, demonstrating its impact on cloud-native storage acceleration for modern AI and big data applications.