Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
This conference talk explores Google Cloud's storage solutions optimized for AI/ML workloads, presented by Marco Abella, Product Manager at Google Cloud Storage. Learn about the critical role storage plays in AI pipelines and how inadequate storage can bottleneck GPU utilization. Discover two industry-optimized storage types: object storage (Cloud Storage) for persistent, high-throughput storage with virtually unlimited capacity, and parallel file systems (Managed Lustre) for ultra-low latency applications. Understand typical AI/ML storage requirements including vast capacity, high aggregate throughput, millions of requests per second, and low-latency reads across different training profiles. Explore Cloud Storage Fuse, which enables mounting a bucket as a local file system with features like file cache, parallel download, streaming writes, and Hierarchical Namespace bucket integration - delivering up to 9x faster model loading times than FSSpec and 30x faster checkpointing performance. Get introduced to Anywhere Cache, a GA feature that improves performance by co-locating storage on SSD in the same zone as compute, reducing time to first byte latency by up to 70% for regional buckets and 96% for multi-regional buckets without requiring code refactoring. Examine a GenAI customer case study demonstrating 99% cache hit rates and reduced network egress costs. Recorded live in Santa Clara on April 22, 2025, as part of AI Infrastructure Field Day.
