Cognitive and SelfAdaptive System for Distributed Tracing
Lightningat ContainerCon @ Open Source Summit Europe 2024 · on Monday, 2024-09-16· at 14:45· Room 0.14 (Level 0)·
summary
Built an adaptive sampling layer over Jaeger and Zipkin that shifts from 1–5% random head sampling to tail-based sampling using semi-supervised density clustering, ingesting spans via Kafka ingestion and a span aggregator. Feature vectors include operation, path, service, host, status code and duration; clusters get inverse weights so rare high-interest traces are retained. The prototype is implemented and reportedly cuts storage needs, lowers tracing costs and significantly improves MTTR for SRE teams.
AI-written summary · about
chapters (6)
- 0:00Distributed tracing intro
- 1:51Tracing components basics
- 3:39Traditional sampling problems
- 5:45Tail-based adaptive sampling
- 9:01Adaptive sampler engine
- 10:53Q&A
related talks
11mCognitive and Self-Adaptive System for Effective Distributed-Tracing...
14mLightning Talk: Cognitive and SelfAdaptive System for Distributed Tracing
13mCognitive and Self-Adaptive System for Effective Distributed-Tracing
13mDistributed Tracing: Adaptive and Telemetry-Based Approach for Effective...
24mData analytics with distributed tracing data
49mDistributed Tracing: How the Pros Debug Concurrent and Distributed Systems
45mMastering Distributed Tracing: Insights from Building a Platform from Scratch
31mDistributed Application Tracing with Trace4Cats