From Continuous Tracing To Automated Insights: AI-Driven Performance Analysis With Guider
summary
Guider is a Python-based profiler that runs as a low-overhead daemon, collects kernel/user data via BPF, perf, ptrace, cgroups and Android hooks, and 200+ subcommands support ad‑hoc and automated workflows. The daemon generates alerts and daemon generates retroactive reports (JSON snapshots plus SVG timelines) when thresholds fire; examples show ANR-trigger windows of 5s plus an extra ~6s delay and a case where 31 of 32 binder threads stall. An LLM pipeline ingests those structured reports to classify root cause and can trigger whitelisted remediation commands whitelist for auto-commands (experimental in field use).
AI-written summary · about
chapters (7)
- 0:00Introduction
- 2:07Guide development and features
- 6:28Daemon and report generation
- 11:04AI-based automated analysis
- 18:01QNX platform support and cases
- 23:42Performance analysis with Guide data
- 36:12Q&A
related talks
39mGuider: A New Performance Monitoring Daemon
53mGuider: Performance Tracing Features for Linux
20mGuider: Lightweight Real-Time Performance & Fault Monitoring Framework for Embedded...
36mIntroducing a New Performance Monitoring Daemon
40mDebugging Android Performance at Scale: Lessons from Building Datadog's Mobile Profiler
35mIn Search of Performance: the Subtle Art of Android Profiling
40mKeep AI on Track: Guardrails + OpenTelemetry Observability
52mDevtools, AI, Insights