How do you make sense of 300 billion distributed tracing spans per day? At Salesforce, the Monitoring Cloud Telemetry Tracer team tackles this challenge head-on — using Apache Flink to process massive real-time telemetry streams and construct accurate, up-to-date service dependency maps across hundreds of microservices.

In this session, we’ll share key architectural decisions, scaling lessons, and operational insights from running telemetry pipelines at extreme scale. You’ll learn how we:

• Process and correlate hundreds of billions of spans in real time
• Design robust stateful streaming pipelines for telemetry data
• Handle out-of-order events and massive fan-out scenarios
• Manage partitioning and state management at scale for high-throughput workloads
• Maintain performance and reliability in mission-critical observability systems

This talk is a must-watch for engineers and architects building large-scale telemetry, tracing, or observability pipelines — or anyone interested in pushing Flink to its real-time processing limits.