Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore how to build and deploy a massive-scale, cloud-agnostic elastic file system designed specifically for autonomous driving research and development in this 31-minute conference talk from the Linux Foundation. Learn about the unique storage challenges faced by autonomous vehicle teams who generate petabytes of multi-modal sensor, simulation, and annotation data monthly, and discover a production-ready open-source solution already implemented at multiple OEMs and robo-taxi startups. Examine the architecture of a file system that supports over 100 petabytes and 100 billion files within a single namespace, backed by commodity object storage while maintaining POSIX compatibility for seamless integration with existing tools. Understand how thousands of Kubernetes nodes with mixed GPU and CPU configurations can share identical datasets for training, validation, and replay operations without data duplication. Discover the multi-protocol endpoint capabilities including POSIX, S3, HDFS, and WebDAV that enable computer vision, mapping, and simulation teams to maintain their current workflows without disruption. Analyze hybrid and multi-cloud deployment strategies that work across public cloud environments and on-premises infrastructure with Ceph, supporting fully air-gapped software-defined vehicle laboratories. Dive into key design patterns including stateless clients, scale-out metadata management, and transparent tiering systems that enable this level of performance and scalability. Review real-world implementation results demonstrating millions in storage cost savings, 10x increases in single-volume capacity, and seamless vendor portability, while gaining access to practical reference diagrams for immediate implementation in your own autonomous driving development environment.
