Resource Management WG Deep Dive
summary
Kubernetes Resource Management WG explains concrete outcomes: it ships device plugins (beta) with production GPU support and hot‑pluggable driver bundles; adds scheduler extenders and node-level device plugins so new resources plug in without core changes; implements static CPU pinning and hugepage support for latency‑sensitive workloads; commits to hardware topology awareness and cluster‑level resources; grows to 30+ active contributors and ships much of this in about eight months with real production users.
AI-written summary · about
chapters (10)
- 0:00Start
- 2:17Improve Compute Resource Management
- 5:49Extensibility
- 6:19Key Elements
- 7:05Scheduler Extensions
- 16:19Application Performance Benchmarking
- 26:38Cpu Quota
- 27:08Disabling Cpu Quota by Default
- 30:59Tuning Cpu Quota Period
- 34:03Would It Make Sense To Have a Single Unified Extension Point across all Different Kinds of Devices like Storage Network and Compute
related talks
30mKubernetes WG Device Management - GPUs, TPUs, NICs and More With DRA
42mGuide Us on Next Steps in Kubernetes Device Manage...
36mTaming the Beast: Advanced Resource Management With Kubernetes
26mResource as You Wish: Collaborative Reservation and Allocation b...
50mDynamic Resource Allocation - the path towards GA
40mBuilding a Fine-Grained and Intelligent Resource Management System on Kubernetes
29mWhat Can I Get You? An Introduction to Dynamic Resource Allocation
21mDemystifying Resource Management in Kubernetes