Nutanix AI Stack Architecture and Components
Nutanix AI Stack Architecture and Components
This domain tests whether you can place each AI capability in the correct Nutanix stack layer and select the right service for a workload, not merely name products.
Core layers
- Nutanix Cloud Infrastructure (NCI): AOS distributed storage, AHV hypervisor, and the Prism Element/Prism Central management plane.
- Nutanix Kubernetes Platform (NKP): CNCF-conformant Kubernetes lifecycle management for containerized AI.
- Data services: Files (NFS/SMB), Objects (S3-compatible), Volumes (block), and Nutanix Database Service (NDB).
- AI-ready blueprint: Nutanix GPT-in-a-Box packages validated compute, GPU, Kubernetes, and model-serving tooling; it is a reference architecture, not a model or SaaS.
- Operations and security: Prism Central, Nutanix Cloud Manager (NCM), Flow Network Security, Life Cycle Manager (LCM), and NCC.
Worked example
A team needs a shared training dataset, containerized training jobs, and a PostgreSQL metadata store. Map the dataset to Files or Objects, training to NKP with GPU node pools on AHV, and metadata to NDB-managed PostgreSQL. Reason from access protocol and lifecycle, not product familiarity.
Traps
- Calling GPT-in-a-Box a model or a cloud service.
- Assuming Objects is a full drop-in for every S3 API feature.
- Placing persistent AI data on ephemeral container storage.
- Forgetting that Prism Central is the cross-cluster management and policy plane.