What does an endpoint give you that a single deployment does not?
Why D is the answer
A stable URL that can front several deployments
Microsoft describes an endpoint as a stable, durable URL. Deployments are the resources behind it.
In detail
- An endpoint is a stable URL.
- It can front several deployments, which is what enables safe rollout.
- That stable URL is what makes safe rollout possible.
Use case
An illustrative scenario, not a real incident.
Apps call one stable URL while the team moves traffic between deployments behind it.
See it in different ways
flowchart LR
A["Blue: current"] -->|"90%"| C["Traffic"]
B["Green: new"] -->|"10%"| C
C --> D{"Healthy?"}
D -->|Yes| E["Increase green"]
D -->|No| F["Back to blue"]stateDiagram-v2 [*] --> WhenNeeded WhenNeeded --> Online: during a request WhenNeeded --> Batch: large volume, can wait Online --> Deployments: stable URL, several deployments Batch --> [*] Deployments --> [*]
sequenceDiagram participant P0 as App participant P1 as Endpoint participant P2 as Deployment A participant P3 as Deployment B P0->>P1: Scoring request P1->>P2: Most traffic P1->>P3: Some traffic
Key point: An endpoint fronts several deployments.
Why the other options are wrong
- A. A guarantee that the model will not drift over time
- Nothing can guarantee that.
- B. Automatic retraining whenever accuracy gets worse
- Retraining is a separate process.
- C. Free compute for the first deployment in a workspace
- Compute is not free.

