Kubernetes for ML — key patterns.
mediumAnswer
- (1) Deployment + Service for online serving.
- (2) Job / CronJob for batch training / scoring.
- (3) StatefulSet for stateful (feature stores).
- (4) HPA for autoscaling.
- (5) NodePool with GPU labels + taints (tolerations for GPU workloads).
- (6) Volumes for shared data (S3-CSI, FSx, NFS).
- (7) Priority + preemption for expensive jobs.
- (8) Namespaces for isolation.
- Operators: Kubeflow, Ray, Volcano scheduler for batch.
- Cost tools: Kubecost.
- Standard for enterprise ML.
Check yourself — multiple choice
- Random
- Deployment+Service online + Job/CronJob batch + StatefulSet FS + HPA autoscale + GPU node pool with taints/tolerations + volumes (S3-CSI/FSx/NFS) + priority + namespaces; Kubeflow/Ray/Volcano operators; Kubecost
- Just Pod
- Not real
K8s for ML: Deploy/Job/StatefulSet + HPA + GPU pool + volumes + operators.
#infrastructure
Practise MLOps & Data Quality
215 interview questions in this topic.
Related questions
- Online vs offline feature store — architecture.
- How do you compute real-time features (streaming aggregates)?
- Orchestration tools for ML pipelines — Airflow vs Prefect vs Dagster.
- Why partition ML training tables by date?
- What is an 'asset' in Dagster and why is it useful for ML?
- Push vs pull materialization for features.