“Cloud Native” OpenCue
summary
OpenCue runs on Google Kubernetes Engine with render pods per renderer, Terraform + Helm/Helmfile for infra, Prometheus for metrics and KEDA-driven scaling to allow node pools to scale to zero. A Python openq controller sets shutdownCost annotations and adjusts KEDA limits; admission webhook ensures stable hostnames. All nodes use spot/preemptible instances; automated restarts recover long renders though preemptions force reschedules. Project has ~15 contributing companies and ~44 individual contributors and is deployed in production.
AI-written summary · about
chapters (12)
- 0:00Introduction to OpenCue
- 1:57Cloud strategy at beloFX
- 4:29Defining render requirements
- 6:11Selecting OpenCue
- 7:51OpenCue architecture recap
- 9:21Initial cloud implementation
- 12:10Addressing scaling challenges
- 16:42Revised cloud infrastructure
- 21:51Infrastructure and tooling
- 24:00Compute node configuration
- 24:54Future development goals
- 27:20Conclusions and resources
related talks
22mAnatomy of a production grade Kubernetes cluster
18mA Cloud-Native Tool for the mass
38mBeyond Kubernetes: Adapting To Spe...
19mKeynote: Building Bridges: Cloud Native and High Performance Computing
26mCloud kubernetes Native Applications
09mKeynote: Build an Open Source Distributed Cloud Native World
21mKeynote: CERN Experiences - Ricardo Rocha & Clenimar Filemon
40mHarnessing the Power of Open Source for Cloud Native Developm...