Question 320
A project contributor checks the outcome after addressing this problem: A model serving design ignores GPU failover because its processors are powerful enough to handle nominal demand. Which result most directly confirms that the intended compute and AI workflow is working?
Explanation
Correct Answer
Service tests meet defined response targets after a supported node failure. This outcome validates the targeted correction: Model fault-domain capacity and appropriate service redundancy for required availability. The original fault mechanism was: Peak performance does not address loss of a host or accelerator resource.
Incorrect Answers
This measurement is relevant when model load time and repeated I/O exceed the available storage path performance. It cannot verify that the reported problem is fixed. Service tests meet defined response targets after a supported node failure.
Peak performance does not address loss of a host or accelerator resource. The selected observation would assess a different failure: Single network components create avoidable availability risks for a multiuser AI service. Evidence for the current fix is: Service tests meet defined response targets after a supported node failure.
NIM availability and support depend on the model, container and approved runtime combination. That explains the usefulness of the selected check, but the outcome required here is different. Peak performance does not address loss of a host or accelerator resource. Service tests meet defined response targets after a supported node failure.