Intro + Deep Dive: Kubernetes IoT Edge WG
summary
Three reference architectures are compared: whole k3s clusters at the edge, cloud control plane with lightweight kubeedge agents (agent ~10 MB), and hierarchical clusters using virtual kubelet. Resource recommendations: always set CPU/memory requests, limits, priorities and QoS to avoid OOM kills and noisy neighbors; Kubernetes scales to ~5,000 nodes but edge networks are intermittent. Data-plane solution uses Qpid Dispatch Router + Apache ActiveMQ Artemis and AMQP 1.0 to route messages across OpenShift clusters without VPNs; demo shows local handling and global load offload working.
AI-written summary · about
chapters (30)
- 0:00Speakers
- 4:28Common requirements for edge workloads
- 4:45Edge types
- 6:11Available Approaches
- 7:36Standard Kubernetes Architecture
- 8:55whole clusters at edge
- 9:57central control managing edge
- 11:19hierarchical edge architecture
- 12:14Comparisons
- 14:01Issues Unique to Edge
- 15:24Edge challenges - recap
- 16:16A small mistake, big consequences
- 19:55Deeper Dive into Admission & Enforcement
- 21:06Ful specs always better
- 24:53Network Policy
- 26:31Bandwidth plugin
- 27:21Example bandwidth Pod spec
- 27:35Audience Poll / Break
- 30:27TCP Oriented Approach
- 31:46AMQP 1.0 Features
- 33:09Routing vs Brokering
- 35:14AMQP 1.0 Infrastructure Components
- 35:54Router Based Multi Region Message Bus
- 36:35Secure multi-site communication
- 47:07Material used in demo is here
- 49:55Edge Environment Differences
- 53:08Edge Security Challenges - Overview
- 53:49Trusting Edge Hardware
- 57:36Trusting Connected Devices
- 59:59Operating System
related talks
34mKubernetes IoT and Edge Working Group
34mKubernetes IoT and Edge Working Group
20mKubernetes at the Edge
34mManaging Edge Nodes With “KubeEdge”
39mShould You Bring Kubernetes on Your Edge Roadtrip?
36mEdge Computing in Kubernetes: Best Practices for the Future
36mEdge Computing and Kubernetes BOF
01h 25mIntro + Deep Dive: Specialized Network Protocols for IoT+Edge with Kubernetes