Improved TiKV Observability: How We Trace Events under Nanoseconds Latency
Talkat KubeCon + CloudNativeCon North America 2020 Virtual · on Tuesday, 2020-11-17· at 15:40· Intrado Virtual Event Platform,·
summary
TiKV adds a custom OpenTracing-compatible tracer (mini-trace) that records every span with only a few nanoseconds overhead, using thread-local buffers to avoid global contention, RDTSCP TSC reads for sub-microsecond timestamps, and latency-based selection so only slow requests ship to collectors. Microbenchmarks show 0.02 microseconds for span gen+collect, 17.5× faster than was-tracing and 100× faster than tokio-tracing, and fallbacks when TSC is unreliable; implementations are open source in Rust and Go.
AI-written summary · about
chapters (3)
related talks
38mLTTng: Beyond Ring-Buffer Based Tracing
27mBeyond OTLP: Unlocking the Potential of OS-native Tracing
37mKernel and userspace tracing with LTTng and friends
48mEfficient and Large-scale Infrastructure Monitoring With Tracing
38mPractical introduction to OpenTelemetry tracing
36mOpenTracing Project Deep Dive
32mDistributed Tracing in Action: Our Journey with OpenTelemetry
45mMastering Distributed Tracing: Insights from Building a Platform from Scratch