Key Definitions
This section provides definitions for key terms and concepts used throughout the Armada Bridge documentation.
| Term | Definition |
|---|---|
| AI Studio | Platform to create and deploy AI/ML applications and models |
| API Key | Authentication key for running API commands |
| Applications | AI/ML and GenAI models and applications |
| Bare Metal | A Server that is allocatable to a tenant, with all its resources (networking, storage etc.) |
| Cluster Template | A template to create Kubernetes or Slurm clusters for running applications |
| CNI Plugin | Plugin for a network interface for Kubernetes clusters |
| Endpoint | HTTP/HTTPS end-point for accessing AI/ML or GenAI models |
| Hugging Face | Model repository |
| JupyterLab | Workspace for running Jupyter notebooks, which enables development and fine tuning of AI/ML applications |
| Kubernetes Cluster | A cluster of nodes created using Kubernetes |
| MiG | Multi-instance GPU, which is NVIDIA technology to divide a single GPU into multiple instances that are isolated from each other |
| Model (AI/ML or GenAI) | In this context, it refers to AI/ML model that is created for specific applications |
| NIM | NVIDIA Inference models which is a repository of curated AI/ML models from NVIDIA |
| Object Storage/S3 | High capacity storage with S3 interface, which is typically used as a tier-2 storage |
| Parallel File System | High performance file system with a cluster of nodes |
| Security Group | A group created with specific security permissions |
| SLURM | Simple Linux Utility for Resource Management, a clustering solution used for HPC (High performance computing) workloads |
| Tenant | An Organization (Enterprise or a Department within an Enterprise) with a set of users |
| Tenant Admin | An Administrator in a Tenant organization (typically an IT admin in an Enterprise) |
| Tenant User | An end user or the consumer of GPU resources in a Tenant organization |
| Token | Authentication token for accessing services |
| End User | Individual account with access to the platform |
| vGPU | NVIDIA technology that allows a GPU or a set of GPUs to be shared among multiple users in a time-shared manner |
| Virtual Machine | A compute node that is carved out of a single physical server into multiple virtual servers |
| VPC | Virtual private cloud, which is a private network for a single tenant in a multi-tenant environment |