Is There a Place for Performance Sensitive Workloads in Kubernetes?
Talkat KubeCon + CloudNativeCon North America 2019 · on Wednesday, 2019-11-20· at 15:20· Room 15AB - San Diego Convention Center, Mezzanine Level·
summary
Examines gaps in Kubernetes for telecommunications-grade, real-time, performance-sensitive workloads (edge/Radio Edge Cloud) and identifies specific architectural shortcomings. Describes existing and proposed open-source technologies and projects that can be combined to build a production-grade, Kubernetes-based edge infrastructure, referencing Akraino Radio Edge Cloud as an example.
AI-written summary · about
chapters (24)
- 0:00Intro
- 0:54What do we do?
- 2:45Open source entanglements
- 4:53Lessons Learned & Gaps
- 7:29"Solutions" out-of-scope and question
- 8:36Realtime Scorecard for K8s 1.16
- 9:38CPU management baseline
- 12:10CPU management - Need moar pools!
- 16:06CPU management - CPU-Pooler
- 16:31CPU management - Pooler in actions
- 17:33CPU management enhanced
- 19:13HW acceleration baseline
- 20:32HW acceleration devices
- 21:30HW acceleration enhanced
- 22:22Memory management baseline
- 23:38Memory management enhanced
- 25:05Network management baseline
- 26:17Network management requirements in production
- 29:32Network management eco-system
- 30:30Networking enhanced
- 32:24Don't believe the hype
- 33:01Topology awareness baseline
- 34:18Topology awareness enhanced
- 36:45References
related talks
20mKubernetes at the Edge for Telcos
39mShould You Bring Kubernetes on Your Edge Roadtrip?
01h 16mIntro + Deep Dive: Kubernetes IoT Edge WG
01h 01mDo We Really Need Kubernetes at the Edge
23mMigrating 5G Core to the Edge, Wisely and Confidently with Kubernetes
36mEdge Computing in Kubernetes: Best Practices for the Future
09mSharpen Your Edge Three Ways With Kubernetes
36mEdge Computing and Kubernetes BOF