Ray Observability - Present and Future

Explore the current and future landscape of Ray observability in this 32-minute conference talk from Ray Summit 2022. Dive into Ray 2.0's observability architecture, learning how to view metrics, logs, and inspect the state of tasks, actors, and other resources. Discover features like task timeline and tracing, and gain insights into the roadmap for Ray observability's future, including a unified observability data model. Ideal for those interested in debugging Ray programs or curious about implementing observability in general-purpose distributed systems like Ray, this talk addresses the challenges of providing a common architecture for both ML offline and online serving users who utilize Ray in vastly different ways.