PMLE Practice Question: Collaborating Within and Across Teams to Manage Data and Models

Your team owns a Vertex AI Model Registry entry that other teams depend on for production serving. A new retrained model shows better offline metrics, and you need to roll it out gradually to a small percentage of live traffic while keeping the ability to revert instantly if quality degrades. What should you do?

⚠ Common exam trap

Many candidates confuse version promotion via aliases with traffic management, when only an endpoint traffic split gives gradual exposure and instant rollback.

Answer choices

Why each option matters

Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.

Correct answer & explanation

✓

Deploy the new model version to the existing endpoint with a traffic split, then shift the split percentage as confidence grows.

Gradual rollout with instant rollback on Vertex AI means deploying both model versions to the same endpoint and controlling the percentage of traffic each receives. Starting with a small split limits blast radius, and because the prior version stays deployed, reverting is a split change rather than a redeployment. Alias promotion, duplicate endpoints, and deleting the old version all fail to provide controlled, reversible traffic shifting.

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

✗
Create a second endpoint for the new model and update the client application to call both endpoints.
Why it's wrong here
Two separate endpoints duplicate infrastructure cost and push traffic-splitting logic into every client, which is brittle and inconsistent. It also provides no unified revert path and complicates monitoring because metrics are split across endpoints. The scenario calls for managed traffic control on one endpoint, not client-side fan-out.
✗
Register the new model under a new model resource and delete the previous version to avoid ambiguity.
Why it's wrong here
Deleting the previous version removes the ability to revert instantly, which is the core safety requirement. It also breaks anything referencing that version and destroys the version history other teams depend on. Creating a separate resource without a traffic split further fragments governance, so this approach is both risky and incorrect.
✗
Assign the new model version the default alias in Model Registry and redeploy the endpoint from that alias.
Why it's wrong here
Changing the default alias affects which version new deployments resolve to, but it does not by itself create a gradual traffic shift on a live endpoint, and it offers no built-in percentage control. This is a promotion mechanism, not a canary mechanism, so it cannot deliver the incremental rollout with instant reversion the scenario requires.
✓
Deploy the new model version to the existing endpoint with a traffic split, then shift the split percentage as confidence grows.
Why this is correct
Vertex AI Endpoints support deploying multiple model versions and splitting prediction traffic by percentage. Deploying the new version alongside the current one and starting with a small split enables a controlled canary rollout, and because the old version remains deployed, reverting is simply resetting the split back to the previous version.

Quick reference

AAA Protocol Comparison

Protocol	Port(s)	Encryption	Transport	Primary Use
RADIUS	1812 / 1813	Password only	UDP	Network access control
TACACS+	49	Full packet	TCP	Device administration
Diameter	3868	Full session	TCP / SCTP	Carrier / mobile networks
802.1X	—	EAP-based	Layer 2	Port-based access control

TACACS+ encrypts the entire packet; RADIUS only encrypts the password field — a key exam distinction.

Go deeper

PMLE Practice Question: Collaborating Within and Across Teams to Manage Data and Models

Why each option matters

Option-by-option breakdown

AAA Protocol Comparison

Related to this question