Storage in the Era of Large-scale AI Computing

Explore a comprehensive conference talk from SNIA Storage Developer Conference 2024 that delves into the critical challenges and solutions for storage systems in large-scale AI computing environments. Learn about the growing performance gap between compute and storage capabilities, particularly in AI workloads, while examining the diverse storage requirements including cost-effectiveness, availability, reliability, IOPS, throughput, and security. Gain insights into the unique storage demands of emerging AI applications like Vector DBs and RAGs compared to traditional AI training and inference workloads. Discover how to address the fundamental challenge of reducing data-entropy across networks while maintaining balanced IO processing. Through AMD presenter Pratik Mishra's expertise, understand the full-stack infrastructure view of AI's impact on computing layers, explore existing optimization solutions, and learn how to adapt current storage systems to meet AI's evolving demands. Master the core challenges in AI computing infrastructure, identify specific storage requirements for AI-focused data centers, and develop strategies to implement existing solutions for AI storage optimization.