Question 50
Evermere Health processes care team access and regulated data handling with controlled access to customer records. The investigation finds: A GPU-based machine learning preparation task uses CPU-only workers despite a tested requirement for GPU acceleration. The team reports that the correction is complete. Which result BEST verifies that the specific issue was resolved?
Explanation
Correct Answer
The test workload detects supported GPU devices and meets its observed performance and cost objective. The observation provides evidence that the proposed fix (Select a compatible GPU-enabled compute configuration after validating driver and runtime prerequisites.) addressed the reported issue (A GPU-based machine learning preparation task uses CPU-only workers despite a tested requirement for GPU acceleration.) for Evermere Health and its care team access and regulated data handling. It verifies an observable outcome instead of relying only on a completed task ticket.
Incorrect Answers
An interruption rehearsal does not cause unrecoverable output loss and the job meets its completion objective. This is suitable evidence for another condition (A retry-sensitive production pipeline uses only preemptible workers and fails repeatedly when available capacity is reclaimed.) after a different response (Use compute capacity appropriate to the job's preemption tolerance and configure supported retries.). For Evermere Health, it does not prove the required result: The test workload detects supported GPU devices and meets its observed performance and cost objective.
Spark task parallelism increases and stage completion improves without exceeding the cost guardrail. This is suitable evidence for another condition (A large parallel Spark pipeline runs on one oversized driver with too few workers, leaving tasks waiting in the execution queue.) after a different response (Rebalance driver and worker configuration and select a supported cluster size suitable for parallel Spark processing.). For Evermere Health, it does not prove the required result: The test workload detects supported GPU devices and meets its observed performance and cost objective.
A disallowed cluster configuration is rejected while an approved cost-bounded configuration can be created. This is suitable evidence for another condition (Workspace users are able to create arbitrary high-cost clusters despite a policy restricting node families and maximum workers.) after a different response (Apply an appropriate compute policy with enforced configuration constraints and permitted settings.). For Evermere Health, it does not prove the required result: The test workload detects supported GPU devices and meets its observed performance and cost objective.