Cognitive and Self-Adaptive System for Effective Distributed-Tracing in Applications

Learn about an innovative machine learning approach to distributed tracing in this 25-minute conference talk from CNCF. Explore how a cognitive and self-adaptive system overcomes traditional limitations in API trace capture by implementing dynamic learning algorithms for unbiased trace collection. Discover how the system moves beyond conventional implementations that typically capture only 5% of normal traces, instead prioritizing diverse trace collection crucial for diagnosing API failures and performance issues. Examine how this ML-based solution streamlines trace metric analysis, enhances reliability work efficiency, and reduces infrastructure costs through targeted trace collection. See how the adaptive sampling approach automatically analyzes system traces and adjusts sampling rates without manual configuration, leading to significant improvements in Mean Time to Resolve (MTTR) and overall operational effectiveness.