Extending Flamegraphs for Multi-Dimensional Performance Analysis
summary
Claims ability to generate flamegraphs from latency, memory, crashdump and kern.log traces and shows concrete results: a memory framegraph that captures an unknown kernel page allocation by sampling 5s every 60s and mapping allocation order to stacks, a latency flamegraph that sorts out low‑latency outliers (identifying a timer interrupt and ru callback hot path), a kronelog/kdump visualization that converts thousands of kernel stacks into an actionable flamegraph, and a prototype easy flamegraph continuous profiler (CPU>80% trigger) plus an MCP server AI workflow that fetches exact kernel source and git commits to triage crashes.
AI-written summary · about
chapters (8)
- 0:00Introduction to multi-dimensional flamegraphs
- 1:04Memory flamegraphs for allocations
- 2:33Latency flamegraphs for outliers
- 4:21Kernel panic analysis with flamegraphs
- 5:24Kdump flamegraphs for crash dumps
- 6:49Ongoing projects overview
- 10:00AI agent for kernel analysis
- 13:11Q&A and feedback discussion
related talks
50mVisualizing Performance - The Developers’ Guide to Flame Graphs
44mNow You See Me Too: Visual Tooling for Advanced System Analysis
33mInto the flamegraph: From the primitives through advanced concepts
56mContinuous Profiling the missing piece in your observability puzzle
44mObservability Tools C++: Beyond GDB and printf - Tools to Understand the Behavior of Your Program
01h 07meBPF: Fueling New Flame Graphs & more
26mProfiling Mastery: Decode Your Data and Elevate Observability
33mA Tale of Two Flamegraphs: Continuous Profiling in Ruby