Managing Edge Computing with Serverless
summary
Kubernetes runs full clusters at each store, using an MQTT broker (Google IoT Core suggested) to push updates and trigger MQTT-triggered rolling upgrades; functions live as container images in an edge registry and use a multi-model database for time series and ML models. The serverless framework demonstrated (Nuclio/nuclear) supports multiple workers per container, GPU inference, shared volumes to avoid copying data, a thumbnailing ML demo, rolling upgrades and rollbacks, and integration with Kubeflow for training in the cloud.
AI-written summary · about
chapters (11)
- 0:00Start
- 2:30Current Approaches are Problematic
- 4:26What Does it take to Move Intelligence to the Edge?
- 5:08Serverless at the intelligent edge
- 6:27But, Serverless comes with challenges
- 8:22Nuclio: taking serverless to the next level
- 18:34Run anywhere, automate data and workload movement
- 21:50Control flow with intelligent edge
- 23:24Simple configuration UI (over MQTT broker)
- 24:22Data movement & analysis flow details
- 26:32Summary
related talks
22mRunning Functions at the Edge
05mContainerization Beyond CPUs – A Kubernetes-based Serverless Pl...
41mIs Serverless the End of Kubernetes?
29mServerless on Kubernetes: What We Learned From Customers
23mServerless with Knative
34mKubeFlow’s Serverless Component: 10x Faster, a 1/10 of the Effort
47mServerless with Knative
21mWhat is the Future of Serverless Computing?