AI-103 Practice Question 31: Provisioned Spillover

Spillover manages exactly this traffic shape: when the provisioned deployment returns a non-200 because PTUs are exhausted, the service automatically routes that request to a standard deployment instead. It needs a provisioned managed deployment and a standard deployment in the same Foundry resource, and is enabled either for the whole deployment (the spilloverDeploymentName property) or per request (the x-ms-spillover-deployment header). Requests the provisioned deployment serves cost only the hourly PTU charge; only the spilled-over requests are billed per token. Microsoft recommends enabling it on all global and data zone provisioned deployments.

- A is wrong because buying a second set of PTUs is the capacity spend finance rejected, and a traffic manager in front of two endpoints is the failover plumbing the team wanted to avoid. - B is wrong because backoff turns a failure into a delay. The request still waits out the burst, and during a flash sale the burst is the entire traffic window. - C is wrong because Global Batch is asynchronous with a 24-hour turnaround, and spillover targets a *standard* deployment. A shopper cannot wait for a batch job.

Reference: Manage traffic with spillover for provisioned deployments