DAMON-Based Pages Migration for CPU, GPU, and XPU Attached and Unattached NUMA Nodes
Linux Plumbers Conference via YouTube
Learn AI, Data Science & Business — Earn Certificates That Get You Hired
PowerBI Data Analyst - Create visualizations and dashboards from scratch
Overview
AI, Data Science & Cloud Certificates from Google, IBM & Meta — 40% Off
One plan covers every Professional Certificate on Coursera. 40% off Coursera Plus Annual.
Unlock All Certificates
Explore advanced memory management techniques in this 19-minute conference talk that examines DAMON-based page migration strategies for complex NUMA topologies featuring CPU, GPU, and accelerator nodes. Learn how the DAMON Linux kernel subsystem monitors data access patterns and enables efficient memory page migration through its DAMOS feature. Discover real-world implementations by major companies including SK Hynix, Micron, and Meta for CXL memory tiering and dynamic memory interleaving solutions. Understand the challenges posed by modern systems with increasingly complex NUMA architectures where nodes may or may not be equipped with various processing units. Gain insights into current DAMON solutions for CXL memory management and future plans for expanding support to broader NUMA topologies. Participate in discussions about core memory management system development, addressing concerns and identifying unmet requirements in the field of non-uniform memory access optimization.
Syllabus
DAMON-based Pages Migration for {C,G,X}PU [un]attached NUMA nodes - SeongJae Park
Taught by
Linux Plumbers Conference