DAMO[N,S]?: Implementing Self-Driven Data Access-Aware Efficient Linux System
summary
DAMON is a kernel subsystem (mainline since v5.15) paired with a Python3 user tool DAMO that produces access heatmaps and drives actions; demo replays show memory footprint dropping from ~10 GB RSS to a 1.8 GB working set when identified cold pages are reclaimed and a live run reduced RSS from ~12 GB to ~6 GB at the cost of slower runtime unless quotas and goals auto‑tune aggressiveness. Overhead is modest (about 2–4% CPU on 100 GB RAM), and DAMO is packaged for pip and multiple distros.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:06Efficient data access importance
- 2:32Monitoring memory access challenges
- 5:07Dynamic granularity for overhead
- 11:31Access heatmaps with demo
- 13:44Live Damon demo
- 19:18Profiling inefficient memory
- 22:39Dynamic time granularity
- 27:45Damos for memory management
- 28:52Damos optimizing memory demo
- 32:31Controlling damos with quotas
- 37:51Performance results trade-offs
- 39:00Damon community future
- 41:27Summary
- 42:21Q&A
related talks
52mDAMON, DAMOS, and DAMO: Kernel Subsystems and User-Space Tools for Data Access...
48mData Access Monitoring Operator (DAMO): User-Space Tool/Python Library for Access-Aware Profiling and Optimization of Your Linux Systems
42mDAMON Recipes: Ways to Save Memory Using a Linux Kernel Subsystem in the Real World- 42mDAMON Recipes: Ways to Save Memory Using a Linux Kernel Subsystem in the Real World
46mDAMON Current Status and Future Plans
45mCurrent Status and Future Plans of DAMON
26mLet's demystify memory
31mHMSDK: Enhancing Software Support for CXL Memory