AI workload storage options

Опубликовано 12 ноября 2025, 23:14

High performance AI and machine learning workloads demand equally high performance storage. This video explores Google Cloud's primary storage recommendations to keep your expensive accelerators fully saturated during both AI training and inference. We dive into Managed Lustre for its unparalleled throughput and Google Cloud Storage (GCS) with GCS FUSE and Anywhere Cache for its flexibility and cost effectiveness. Discover the right storage solution to build the foundation for your AI applications.

Chapters:
0:00 - Introduction
0:31 - Storage for AI Training: Managed Lustre
1:19 - Alternative for Training: Google Cloud Storage with GCS FUSE
2:08 - Storage for Inference: GCS with Anywhere Cache
2:32 - Alternative for Inference: Managed Lustre
2:55 - Summary of Storage Recommendations
3:25 - Conclusion

Resources:
High performance parallel file system → goo.gle/ra-managed-lustre
Optimize AI and ML workloads with Cloud Storage FUSE → goo.gle/ra-gcs-fuse

Subscribe to Google Cloud Tech → goo.gle/GoogleCloudTech

#GoogleCloud #GCSFUSE #CloudStorage

Speakers: Drew Brown
Products Mentioned: AI Infrastructure, Cloud Storage