Sample Your Traffic but Keep the Good Stuff!
summary
Describes concrete sampling recipes: record per-event sample rates, use fields like HTTP status, dataset, method and URL to bias selection, and let servers adapt rates by observed volume so high-volume keys sample more heavily. Implements include Honeytail and Jaeger and nginx/varnish rules; Jaeger uses trace ID hashing to make consistent trace decisions. Claims an order-of-magnitude throughput cut while still seeing rare events, often keeping individual events at one in a thousand overall.
AI-written summary · about
chapters (9)
- 0:00Start
- 3:00Sampling
- 9:31Sample Rate
- 10:25The Static Sample Rate
- 14:19Rate Limiting Sampling
- 20:03Jaeger Distributed Tracing System
- 22:02Api Server
- 22:46Types of Requests
- 26:01Recap
related talks
27mTricked Out Traces
33mReal-World Sampling – Lessons Learned...
14mCognitive and SelfAdaptive System for Distributed Tracing
35mRefining Systems Data without Losing Fidelity
13mDistributed Tracing: Adaptive and Telemetry-Based Approach for Effective...
36mJaeger Deep Dive- 14mLightning Talk: Cognitive and SelfAdaptive System for Distributed Tracing
11mCognitive and Self-Adaptive System for Effective Distributed-Tracing...