The Hitchhiker's Guide to Kubernetes Platforms: Don’t Panic, Just Launch!
Talkat Cloud Native AI Day Europe 2024 · on Tuesday, 2024-03-19· at 13:30· Pavilion 7 | Level 7.3 | Room S03·
summary
Three concrete platform APIs power the system: debugging API that returns a single status summary with customized dashboard links and translated pod errors, deployment API that takes YAML manifests, enforces namespace RBAC, checks CPU/GPU quotas and deploys to a paired multi-cluster DR setup, and versioning API backed by CRDs stored in a space do database to enable rollbacks, diffs and an audit trail. The stack uses kaser for model serving, an internal model registry, resource-insight endpoints, and plans for LLM benchmarking and OpenAI protocol support.
AI-written summary · about
chapters (16)
- 0:00Kubernetes AI platforms intro
- 1:04Six principles for stability
- 2:06KServe for inference
- 3:28Three core AI APIs
- 5:12Debugging API for observability
- 8:19Deployment statuses clarity
- 10:26Error codes translation
- 12:53Deployment API metaphor
- 15:20Multi-tenancy with namespaces
- 17:48Resource allocation checks
- 18:50Disaster recovery clusters
- 21:02Versioning for rollbacks
- 22:28Internal versioning vs GitOps
- 23:11User-driven API design
- 24:14Scalability and LLM support
- 24:57Q&A
related talks
26mKubernetes Is A Platform Platform
20mWhat Does A Platform On Kubernetes Look Like?
08mKubernetes: The API of Everything
19mProduction-Ready Platform Engineering: Build To FAIL
32mKubernetes as a platform for platforms
23mScaling Kubernetes in Regulated Environments Through Platform APIs
42mQuestions answered: Kubernetes Fundamentals and its Advanced Platform Capabilities
31mKubernetes is Your Platform: Design Patterns For Extens...