Simply select "Serverless" when creating a notebook or job
Who this is for:
Architecture / Concept Overview: Simply select "Serverless" when creating a notebook or job
Serverless GPU compute abstracts away infrastructure, providing on-demand GPUs that spin up for training and shut down automatically.
%%{init: {"theme":"base","themeVariables":{"background":"#0B0E14","primaryTextColor":"#E0E6ED","lineColor":"#5D6470","darkMode":true,"primaryColor":"#2E4A4A","secondaryColor":"#374151","secondaryTextColor":"#E0E6ED","tertiaryColor":"#111827","tertiaryTextColor":"#E0E6ED","edgeLabelBackground":"#1f2937"}}}%%
flowchart LR
classDef source fill:#3F4B59,stroke:#9CA3AF,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef ingestion fill:#5A4B36,stroke:#C9A86B,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef processing fill:#535072,stroke:#8E82B4,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef storage fill:#2E4A4A,stroke:#5FAFA8,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef serving fill:#3D5550,stroke:#6BB7AA,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef governance fill:#5A3F52,stroke:#C28BB0,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
USER[Data Scientist] -->|Submit| JOB[Training Job / Notebook]
JOB -->|Request| POOL[Serverless GPU Pool]
POOL -->|Provision| GPU[GPU Instance]
GPU -->|Train| MODEL[Model Training]
MODEL -->|Log| MLF[MLflow]
GPU -->|Release| POOL
MLF -->|Register| UC[Unity Catalog]
USER:::source
JOB:::ingestion
POOL:::processing
GPU:::storage
MODEL:::serving
MLF:::governance
UC:::governance
*Serverless GPU flow: jobs request GPUs from a managed pool, train models, and release resources automatically.*
%%{init: {"theme":"base","themeVariables":{"background":"#0B0E14","primaryTextColor":"#E0E6ED","lineColor":"#5D6470","darkMode":true,"primaryColor":"#2E4A4A","secondaryColor":"#374151","secondaryTextColor":"#E0E6ED","tertiaryColor":"#111827","tertiaryTextColor":"#E0E6ED","edgeLabelBackground":"#1f2937"}}}%%
graph TD
classDef source fill:#3F4B59,stroke:#9CA3AF,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef ingestion fill:#5A4B36,stroke:#C9A86B,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef processing fill:#535072,stroke:#8E82B4,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef storage fill:#2E4A4A,stroke:#5FAFA8,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef serving fill:#3D5550,stroke:#6BB7AA,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef governance fill:#5A3F52,stroke:#C28BB0,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
CLASSIC[Classic Clusters] --> MANAGE[Manual Provisioning]
CLASSIC --> IDLE[Idle Cost Risk]
CLASSIC --> CONFIG[Complex Configuration]
SERVERLESS[Serverless GPU] --> AUTO[Auto-Provisioning]
SERVERLESS --> PAY[Pay-Per-Use]
SERVERLESS --> SIMPLE[Zero Configuration]
CLASSIC:::source
MANAGE:::source
IDLE:::source
CONFIG:::source
SERVERLESS:::serving
AUTO:::serving
PAY:::serving
SIMPLE:::serving
*Classic clusters vs. serverless GPU: serverless eliminates idle costs and manual configuration.*
Key Terms
Prerequisites and Setup
- Databricks workspace with serverless compute enabled.
- Admin has enabled GPU serverless in workspace settings.
- Sufficient GPU quota in the underlying cloud account.
Step-by-Step Implementation
Configuration Reference
| Parameter | Default | Description |
|---|---|---|
compute_type | Serverless | Select in notebook or job configuration |
gpu_type | auto-selected | GPU instance type (managed by Databricks) |
auto_termination | immediate | Resources released when job completes |
environment_version | latest | Serverless runtime version |