DAMON Recipes: Ways to Save Memory Using a Linux Kernel Subsystem in the Real World
summary
DAMON is a mainline Linux data-access monitor (since v5.15) with userspace tooling that reports hot/cold address ranges and supports virtual and physical modes. Deployments report 3–4% CPU overhead while profiling-driven changes (mlock hot pages under pressure) show up to 2.3× speedup in lab tests. Cloud use enables proactive product reclamation for memory autoscaling (used by AWS variants), and CXL use via SK hynix’s HMS SDK implements demote/promote between DRAM and CXL with measured bounds of 18.8% slowdown (CXL-only) and up to 12.9% improvement when DAMON is active.
AI-written summary · about
chapters (8)
- 0:00DAMON introduction
- 1:03DAMON core features
- 5:35DAMON recipes overview
- 6:38Recipe: Access profiling
- 11:15Recipe: Memory reclamation
- 16:49Recipe: CXL tiering with SK Hynix
- 31:29DAMON community
- 35:48Summary
related talks
42mDAMON Recipes: Ways to Save Memory Using a Linux Kernel Subsystem in the Real World
52mDAMON, DAMOS, and DAMO: Kernel Subsystems and User-Space Tools for Data Access...
45mCurrent Status and Future Plans of DAMON
47mDAMO[N,S]?: Implementing Self-Driven Data Access-Aware Efficient Linux System
46mDAMON Current Status and Future Plans
48mData Access Monitoring Operator (DAMO): User-Space Tool/Python Library for Access-Aware Profiling and Optimization of Your Linux Systems
31mHMSDK: Enhancing Software Support for CXL Memory
43mLive in a world with multiple memory types