Distributed Tracing: Adaptive and Telemetry-Based Approach for Effective...
summary
Claims an adaptive tracing layer that combines head+tail sampling with a density‑based clustering sampler, using feature vectors (operation path, service name, span tags, host, status code) to assign inverse‑frequency weights and sampling thresholds. Built an adaptive sampler that captures all traces for new services briefly, then tunes sampling by cluster, avoiding blind 1–5% random drops and reducing missed fault traces; reports >80% accuracy in surfacing previously missed problematic traces while controlling storage/CPU costs.
AI-written summary · about
chapters (7)
- 0:00Distributed tracing intro
- 1:52Observability in microservices
- 4:20How tracing works
- 6:27Sampling limits in tracing
- 8:15Head vs tail sampling
- 9:40Adaptive sampling with vectors
- 12:49Summary & Q&A
related talks
14mLightning Talk: Cognitive and SelfAdaptive System for Distributed Tracing- 14mCognitive and SelfAdaptive System for Distributed Tracing
13mCognitive and Self-Adaptive System for Effective Distributed-Tracing
11mCognitive and Self-Adaptive System for Effective Distributed-Tracing...
27mTricked Out Traces
26mDistributed Tracing: All the Warning Signs Were Out There!!!
33mReal-World Sampling – Lessons Learned...
34mReduce Complexity with Distributing Tracing