Small Granularity Graph Neural Network Training and the Future of Storage

Explore how Graph Neural Network (GNN) training workloads challenge traditional storage architectures in this 24-minute conference talk from SNIA SDC 2025. Examine the limitations of current SSD devices built with 4KiB transaction units when faced with high IOPS demands at smaller granularities. Dive deep into GNN testing using NVIDIA BaM and discover the modifications made to test sub-4K transactions. Learn how AI/ML training workloads utilize storage devices and understand how these emerging use cases are driving the development of next-generation storage technologies. Gain insights into Storage Next initiatives and discover opportunities to participate in shaping the future of storage development. Understand the intersection of artificial intelligence, machine learning, and storage technology evolution through practical testing examples and real-world performance considerations.