Leveraging Continuous Production Profiling for Providing Insights into...
summary
LinkedIn runs always-on Java profiling using async-profiler in 30-minute chunks, generating ~48 TB raw data per day before compression; retention is three days and on-the-fly compression plus tree/segment indexing shrinks a 400 MB profile to 400 MB → 1.5 MB and cuts load time from >5 minutes to <30 seconds. Automated regex-based tagging separates infra vs app stacks, surfaces regressions (example: CPU hotspot drops from 56% to 19% with a three-line fix), and feeds metrics/alerts for RCA and fleet-level optimizations.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:58How is Profiling at LinkedIn
- 10:31Regex
- 14:26Problems
- 15:50Hardware homogeneity
- 16:41Service Discovery
- 18:05Storage Optimization
- 19:24Open Source Tools
- 19:56Looking Ahead
related talks
40mContinuous Profiling, the missing piece in your observability puzzle!
01h 13mContinuous Profiling: The Missing Piece in Your Observability Puzzle
56mContinuous Profiling the missing piece in your observability puzzle
36mProfiling in the cloud-native era - Matthias Loibl
51mProduction Profiling: What, Why and How
47mContinuous Profiling — what, why and how
46mContinuous Profiling - the 4th Pillar of Observability
52mThe missing piece in your observability puzzle!