DAMON, DAMOS, and DAMO: Kernel Subsystems and User-Space Tools for Data Access...
summary
DAMON is a Linux kernel subsystem (mainline since 5.15/5.16) that implements regions-based sampling and adaptive region tuning to monitor data access with bounded overhead; a production claim is scanning 68 GB every 5 ms while using <1% CPU and identifying a 7 GB working set from 70 GB. DAMOS lets users specify actions (for example DAMOS action: page_out) to apply automatically; evaluations show a 76% THP footprint reduction while preserving 51% of THP speedup, and a reclamation case that reduces resident set by 93% with ~1.2% runtime cost. User tool DAMO and an rmap+PTE-based API enable visualization and automation.
AI-written summary · about
chapters (13)
- 0:00Diamond project introduction
- 3:13Efficient memory management
- 6:03Prior monitoring limits
- 8:11DAMON features
- 12:37DAMON monitoring demo
- 19:36Introducing DAMOS
- 23:03DAMOS demo
- 28:41DAMOS production features
- 35:24User vs kernel monitoring
- 41:01DAMON user tools
- 43:29Putting DAMON, DAMOS, and demo together
- 46:42Diamond future and community
- 48:08Q&A
related talks
48mData Access Monitoring Operator (DAMO): User-Space Tool/Python Library for Access-Aware Profiling and Optimization of Your Linux Systems
47mDAMO[N,S]?: Implementing Self-Driven Data Access-Aware Efficient Linux System
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
17mActionable Data Access Monitoring Output Data and Format
31mHMSDK: Enhancing Software Support for CXL Memory