| Model | Model or AI model is a computer program designed to simulate human intelligence |
| Model Serving | It is a product of Platform as a Service (PaaS) that allows the deployment and operation of AI models on the Kubernetes platform provided by FPT Cloud |
| Model Type | The type of model to be deployed, including/excluding an image |
| Image | A template for packaging applications as containers for deployment |
| Image Source | The type of source for the deployment image: Private/Public |
| Image Registry | The address where the deployment image is hosted |
| Image Tag | The tag of the deployment image |
| Cluster | The Kubernetes cluster for deploying the model |
| Instance | Processing capacity unit |
| Instance Replica | The number of identical instances providing processing capacity for a single model |
| Resource Type Flavor | Type of processing capacity resources (CPU/RAM/DISK/GPU) |
| Deployment Strategy | Deployment strategy on Kubernetes (Recreate/Rolling Update) |
| Startup Command | Startup command or initialization command |
| Environment Variable | Environment variables that need to be configured |
| Nodes Selector | Label selection for worker nodes to deploy the model |
| Tags | Label for the deployed model |
| Services Type | Type of external connection for the deployed model on Kubernetes (Load Balancer/ClusterIP/Ingress) |
| Traffic Type | Type of data connection (Public/Private) |
| Port | Port for external connection and container ingress (TCP) |
| GPU Sharing Mode | The type of virtualization used to partition the GPU (MIG/MPS/Time-slice) |