Extending Kubernetes Scheduler for Multi-Cluster and Multi-Cloud Workloads
summary
Kubernetes API is extended into a multi‑cloud control plane called Crossplane that models AWS, Azure and GCP services as CRDs and controllers, performs continuous reconciliation, and exposes resource classes and claims so apps request a database without knowing provider details. Demonstrated provisioning from a laptop to three clouds (GKE, EKS, AKS) with RDS, Cloud SQL and Azure Database created on demand. Scheduling today uses cluster selectors; Crossplane does not yet sync multi‑cluster dataset replication or distribute a single instance across clusters.
AI-written summary · about
chapters (19)
- 0:00Intro
- 0:37Cloud Computing
- 1:13Cloud Providers
- 2:57Managed Services Overlap
- 3:23Multicloud
- 5:30Kubernetes API
- 6:31Extending Kubernetes
- 7:54Controller
- 8:57Stateful Applications
- 9:56Operator Resources
- 10:29Kubernetes Platform Services
- 12:30External Resources
- 14:27Managed Resources as CRD's
- 15:03Cloud Provider as a Resource
- 15:38Managed Service as a Resource
- 16:55Separation of concerns Application Owner
- 18:13Resource Classes
- 20:50Workload
- 23:00Crossplane Vision
related talks
30mExtending Kubernetes Scheduler for Multi-Cluster and Multi-Cloud Workloads
36mKubernetes as the Universal Control Plane: From Pods to Multi-Cloud Infrastructure
17mCrossplane: Deploy Your Cloud Components from Your K8S Cluster
27mKubernetes as a Universal Infrastructure Control Plane
23mIntroduction to Crossplane and Simplifying Cloud Native Infrastructure Management
25mCrossplane: Build and Manage Infrastructure using Kubernetes
35mBuilding and Enabling APIs with Crossplane 1
26mKubernetes - The Universal Control Plane