Introduction to Managed Lustre with Google Cloud

Explore a comprehensive introduction to Managed Lustre with Google Cloud in this 29-minute tech presentation by Dan Eawaz, Senior Product Manager at Google Cloud. Learn about this fully managed parallel file system built on DDN Exascaler, specifically designed to address the demanding requirements of AI workloads including data preparation, model training, and inference. Discover how Managed Lustre delivers high throughput to maximize GPU and TPU utilization while enabling quick checkpoint writing and reading. The presentation explains how this service helps customers migrate their parallel file system workloads to the cloud without re-architecting, optimizes total cost of ownership, and offers persistent deployment co-located with compute for optimal latency. Get insights into the service's impressive specifications, including scaling from 18 terabytes to petabyte scale, sub-millisecond latency, and initial throughput of one terabyte per second. The talk covers the managed deployment process, 99.9% availability SLA, POSIX compliance, and integrations with GKE, Slurm, and Google Cloud Storage for batch transfers. This presentation was recorded live in Santa Clara, California on April 22, 2025, as part of AI Infrastructure Field Day.