# Intel® Tiber™ AI Studio Installation Requirements
# Kubernetes Requirements
Intel® Tiber™ AI Studio platform can run atop any conformant Kubernetes distribution
Cloud-based distros
- AWS - EKS
- GCP - GKE
- Azure - AKS
On-premise distros
- Kubeadm (1.17+)
- VMware Tanzu Kubernetes Grid
- Rancher (RKE v1, RKE v2)
- OpenShift (v4 +)
- Kubespray (1.17+)
- DeepOps
# Persistent Storage
AI Studio can consume any Kubernetes CSI compatible storage supported and listed in Kubernetes official documents provisioner.
AI Studio Kubernetes Operator can deploy out-of-the-box two storage provisioners. These are suitable for environments where a storage provider isn’t available.
- Hostpath CSI provisioner
- NFS client CSI provisioner
# Kubernetes Ingress Controller
AI Studio can integrate with any supported Kubernetes ingress controller.
AI Studio Kubernetes Operator can deploy out-of-the-box Istio and configure Istio Ingress Gateway to allow ingress traffic into the cluster. Kubernetes Resource Requirement
# Resource Requirements
| CPU request | Memory request | Storage request |
|---|---|---|
| 15CPU | 26GB | 350GB |
NOTE
Please note, these are the minimal defaults, i.e., Kubernetes resources requests. Production resources utilization values will highly depend on the actual load and usage of AI Studio-control-plane
Please note, these compute resources are for AI Studio control plane only. The actual compute power for ML workloads has to be calculated independently and include totals of CPU/GPU, Memory, and storage expected to be consumed by the ML workloads.
# Network Requirements
# Network Subnet and CIDR
AI Studio platform requires a minimum of /25 or lower (recommended /16)
# DNS
AI Studio requires a DNS wildcard subdomain record which resolves the ingress IP to the AI Studio cluster domain. e.g., *.cnvrg.my-org.com -> 1.2.3.4. Istio will allocate the subdomain to the different components of AI Studio as well as for new workspaces and endpoints.
When DNS wildcard subdomain record isn’t available for AI Studio deployment, you might use nip.io
WARNING
Using nip.io is only suitable for PoC or dev environments. And it’s not recommended or supported in production environments
# Required Ports
AI Studio components communicate with worker clusters and on-premise machines, which are communication endpoints that are identifiable for specific processes or services. Ensure the following ports required by AI Studio are open between the Kubernetes environments and or compute machine( i.e., external DGX machine running docker-compose)
| Port | Protocol | Description |
|---|---|---|
| 6443 | TCP | Kubernetes API server |
| 443 | TCP | AI Studio web-UI with SSL |
| 80 | TCP | AI Studio web-UI without SSL |
| 22 | TCP | Required for SSH |
| 25 | TCP | SMTP |
| 5432 | TCP | Required for external PostgreSQL |
| 3000-9000 | TCP | AI Studio notebooks |