5 GB of Traces per Second, or How Ozon's Tracing System Works
summary
Description of Ozon’s request-tracing system that handles 5 GB/s of traces, covering the system architecture, the trace processing pipeline, and storage solutions for large volumes of trace data. The talk also describes operational and development problems encountered while building that high-throughput tracing infrastructure.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:19Microservice architecture problems
- 4:05Distributed tracing
- 5:28Choosing Jaeger
- 7:14Log system integration
- 8:17Critical path tool
- 10:465 GB/s trace load
- 12:32Sampling management
- 15:0015-minute trace storage
- 18:31Jaeger collector architecture
- 20:38Microservice dependency graph
- 22:24Invalid trace filtering
- 24:29Jaeger UI modernization
- 25:53SegDB transition for logs
- 26:56Tag-based trace correlation
- 27:58Client quota mechanism
- 29:03Summary and tips
- 30:06Q&A
related talks
36mA Survey of the OSS Tracing Ecosystem - Ben Sigelman & Ted Young, LightStep
37mObservability, Distributed Tracing & the Complex World
51mHow We Do Tracing at the Scale of Thousands of Services and Millions of Spans per Second
45mMastering Distributed Tracing: Insights from Building a Platform from Scratch
16mFrom Sampling To Full Visibility: Scaling Tracing To Trillions of Spans
29mIntroduction to Distributed Tracing
26mDistributed Tracing FAQ
42mHow Much Is Tracing? Measuring the Overhead Caused by the Tracing Infrastructure