Vulcan: Automatic Query Planning for Live ML Analytics

Explore an innovative system for optimizing live machine learning analytics queries in this 16-minute conference talk from NSDI '24. Dive into Vulcan, an automatic query planning solution that addresses the challenges of manual pipeline construction, configuration, and placement across heterogeneous edge infrastructures. Learn how Vulcan efficiently navigates the vast search space to determine optimal query plans, improving accuracy, latency, and resource consumption. Discover how this system outperforms existing ML analytics solutions, achieving significant improvements in search cost and query latency. Gain insights into Vulcan's ability to perform fast online adaptation post-deployment, making it a powerful tool for large-scale ML deployments. Understand the potential impact of this research on the future of live ML analytics and its applications in various industries.