Challenges in Building Multi-Cloud-Provider Platform With Managed Kubernetes
summary
ArangoDB builds Oasis, a managed multi‑cloud service that runs on GKE AKS and EKS and reports concrete problems: cluster provisioning creates unmanaged cloud resources that must be torn down in order, node sizing and volume IOPS vary wildly so cross‑cloud benchmarks shape SLA choices, and networking needs CNI workarounds (Calico/psyllium) plus per‑cloud setup scripts. Built tooling includes an operator for config rollout and Grafana Loki plus per‑cluster Prometheus; it works but requires per‑provider hacks and constant version monitoring.
AI-written summary · about
chapters (22)
- 0:00Start
- 5:20ArangoDB Oasis
- 5:54Deployment
- 6:47Databases & Container
- 9:57Architecture
- 11:39Controlplane
- 12:10Data Cluster
- 13:09Container Abstractions
- 13:39Container Orchestration
- 16:53Abstraction II (managed)
- 19:55First: Blowing Steam off...
- 23:26Resource Creation
- 25:11Kubernetes Versions
- 25:35Kubernetes Cluster Options
- 28:19Logging/Audit Log
- 29:32Storage
- 31:47Networking for Multi-tenancy
- 32:59Cilium: The Promise
- 34:02Cilium: The Reality
- 35:17On-Prem
- 36:29Container Runtimes
- 38:16Experiences
related talks
33mKubernetes-Driven Multi-Cloud Deployments: Maximizing Scalability and Flexibility
41mAn Adventure in Multi-Cloud with Kubernetes
15mGoing Multi Cloud in a Hurry with Quality and Style
36mThe Architecture of a Multi-Cloud Environment with Kubernetes
30mTransparent Live Migration of Services Between Kubernetes Cluster
34mTaking Back Control: Building Cloud-Agnostic Container Platforms for Regulated Environments
35mKubernetes journey
31mAn Adventure in Multi Cloud with Kubernetes and Distributed Data