Continuous Profiling the missing piece in your observability puzzle
summary
Continuous profiling collects low‑overhead runtime samples (typically 1–2% overhead) and turns them into searchable records and flame graphs so regressions are visible in production instead of only in dev. Expect language profilers (JFR, async-profiler) and kernel‑level eBPF collectors, sampling intervals around 100–200ms, symbolization and safe‑point bias as real problems, and storage/retention costs to grow quickly. Demo workflows show comparing two deployments and finding a CPU spike in the car path in EU‑North via diffed flame charts.
AI-written summary · about
chapters (24)
- 0:00Introduction
- 1:05On-call challenges
- 2:10Monolith to microservices
- 3:55Microservice communication issues
- 6:01Three risk windows in software
- 7:48Observability pillars and limits
- 13:17Profiling unknown unknowns
- 14:44Profiling types and purposes
- 16:34Traditional profiling drawbacks
- 19:24Continuous profiling benefits
- 21:30Continuous profiling architecture
- 24:18Profiling data collection
- 25:00Profiling techniques
- 27:33Profiling challenges
- 30:19Cost savings with profiling
- 32:48Feedback loop and regression
- 35:37Language profilers and safepoints
- 37:27Safepoint bias in Java
- 41:01eBPF introduction
- 46:04eBPF profiling tools
- 49:19Pixie observability platform
- 51:04Demo: Periscope performance
- 54:59eBPF future in observability
- 55:19OpenTelemetry profiling support
related talks
52mThe missing piece in your observability puzzle!
01h 13mContinuous Profiling: The Missing Piece in Your Observability Puzzle
40mContinuous Profiling, the missing piece in your observability puzzle!
24mCommunity Driven Breakout Session: Continuous Profiling
46mContinuous Profiling - the 4th Pillar of Observability
24mWhy Does Continuous Profiling Matter to Developers?
36mContinuous Profiling in the Cloud-Native era
26mProfiling Mastery: Decode Your Data and Elevate Observability