Introducing In-Kernel PSI Auto Monitor Feature
Talkat Open Source Summit India 2026 · on Wednesday, 2026-06-17· at 17:15· Lotus Ballroom 3 (Level 3)·
summary
Optional in-kernel PSI Auto Monitor captures thread-level contributors exactly when configured PSI thresholds breach, using existing PSI tracepoints and a workqueue so there are no changes to PSI fast paths and no always-on daemon. It sorts and dumps top N contending tasks (demo used top 20, 14 selected), runs during early boot, reports trigger latency and low overhead across kernels 5.4–7.0 and PREEMPT_RT, and pairs with a GenAI pipeline that parses logs into timelines and actionable summaries.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:28What is PSI in Linux
- 3:12Problem statement
- 6:00Existing PSI solutions
- 7:26Introducing PSI auto monitor
- 10:38PSI auto monitor thresholds
- 11:20Results on BeagleBone
- 14:30Comparison with top
- 15:34Early boot-time monitoring
- 17:20Tracing interface options
- 17:40Gen AI log analysis
- 20:30Testing with real-time kernel
- 21:53Key benefits
- 23:21Upstream plans
- 26:05Summary
- 28:52Q&A
related talks
35mIntegrated PCIe monitoring and tracing facilities
59mProposing a new tracer to monitor RT task behavior
44mAdvanced Linux Kernel Networking Monitoring
45mLinux perf tool metrics
45mFast and Safe Production Monitoring of JVM Applications with BPF Magic
52mEffectively Measure and Reduce Kernel Latencies for Real-time Constraints
13mEnabling DDR segments on demand during memory pressure for DDR power reduction
36mObservability and performance analysis with BPF