Exam Bundle

This signal belongs to the adjacent case “A very small table receives no automatic clustering keys despite configuration,” whose remedy is Accept cost-aware optimizer selection rather than force arbitrary keys. The observed result for the present requirement is Automatic/manual maintenance targets tables with demonstrated query benefit.

That observation is a valid indicator for “The team wants Databricks to manage table data lifecycle and optimization automatically” when using Create a Unity Catalog managed Delta table with eligible predictive optimization. The scenario instead tests “Maintenance cost is rising on tables rarely queried”; the relevant result is Automatic/manual maintenance targets tables with demonstrated query benefit.

File count and average file size improve in that table pertains to a different control (Execute a scoped OPTIMIZE maintenance job during a suitable window) and a different need (One unusually fragmented table needs a controlled immediate compaction). It cannot establish that Use cost-aware optimization policies rather than blanket frequent rewrites met the stated goal.

Persist the justified reused intermediate and release after use is appropriate when “Repeated recomputation of an intermediate DataFrame dominates a notebook”; it would be verified by Subsequent actions avoid rerunning the expensive ancestor plan. It does not directly satisfy “The report reads more files than before despite unchanged filter semantics,” whose relevant acceptance evidence is Scan/file metrics identify the additional read volume.

Measure startup separately and change supported warehouse readiness settings belongs to the separate case “The job is spending most time waiting for a cold warehouse to start,” where Cold-start component of end-to-end latency decreases would be expected. That differs from the condition in this stem: “The report reads more files than before despite unchanged filter semantics.”

Run result-equivalence and data-quality regression before accepting the optimization belongs to the separate case “A rewrite made the job faster but changes the number of output records,” where The final optimized output agrees with the validated baseline would be expected. That differs from the condition in this stem: “The report reads more files than before despite unchanged filter semantics.”

Straggler duration falls while output counts stay consistent would demonstrate the separate action Inspect skew metrics and apply AQE/skew mitigation, which addresses “The report's join stage contains one extreme task that dominates runtime.” It does not verify the requirement “The report reads more files than before despite unchanged filter semantics” following Inspect file pruning, statistics and changed data layout.

That observation is a valid indicator for “The job is spending most time waiting for a cold warehouse to start” when using Measure startup separately and change supported warehouse readiness settings. The scenario instead tests “The report reads more files than before despite unchanged filter semantics”; the relevant result is Scan/file metrics identify the additional read volume.

That observation is a valid indicator for “Repeated recomputation of an intermediate DataFrame dominates a notebook” when using Persist the justified reused intermediate and release after use. The scenario instead tests “The report reads more files than before despite unchanged filter semantics”; the relevant result is Scan/file metrics identify the additional read volume.

The alternative Run result-equivalence and data-quality regression before accepting the optimization has a distinct purpose—A rewrite made the job faster but changes the number of output records. It would not establish the required outcome (The report's join stage contains one extreme task that dominates runtime); the necessary verification concerns Straggler duration falls while output counts stay consistent.

Measure startup separately and change supported warehouse readiness settings belongs to the separate case “The job is spending most time waiting for a cold warehouse to start,” where Cold-start component of end-to-end latency decreases would be expected. That differs from the condition in this stem: “The report's join stage contains one extreme task that dominates runtime.”

The alternative Persist the justified reused intermediate and release after use has a distinct purpose—Repeated recomputation of an intermediate DataFrame dominates a notebook. It would not establish the required outcome (The report's join stage contains one extreme task that dominates runtime); the necessary verification concerns Straggler duration falls while output counts stay consistent.

Scan/file metrics identify the additional read volume would demonstrate the separate action Inspect file pruning, statistics and changed data layout, which addresses “The report reads more files than before despite unchanged filter semantics.” It does not verify the requirement “The report's join stage contains one extreme task that dominates runtime” following Inspect skew metrics and apply AQE/skew mitigation.

Subsequent actions avoid rerunning the expensive ancestor plan pertains to a different control (Persist the justified reused intermediate and release after use) and a different need (Repeated recomputation of an intermediate DataFrame dominates a notebook). It cannot establish that Inspect skew metrics and apply AQE/skew mitigation met the stated goal.

Cold-start component of end-to-end latency decreases would demonstrate the separate action Measure startup separately and change supported warehouse readiness settings, which addresses “The job is spending most time waiting for a cold warehouse to start.” It does not verify the requirement “The report's join stage contains one extreme task that dominates runtime” following Inspect skew metrics and apply AQE/skew mitigation.

The alternative Inspect skew metrics and apply AQE/skew mitigation has a distinct purpose—The report's join stage contains one extreme task that dominates runtime. It would not establish the required outcome (The job is spending most time waiting for a cold warehouse to start); the necessary verification concerns Cold-start component of end-to-end latency decreases.

Persist the justified reused intermediate and release after use is appropriate when “Repeated recomputation of an intermediate DataFrame dominates a notebook”; it would be verified by Subsequent actions avoid rerunning the expensive ancestor plan. It does not directly satisfy “The job is spending most time waiting for a cold warehouse to start,” whose relevant acceptance evidence is Cold-start component of end-to-end latency decreases.

This action solves a different issue: “A rewrite made the job faster but changes the number of output records.” Its normal signal is The final optimized output agrees with the validated baseline. Here the requested behavior is “The job is spending most time waiting for a cold warehouse to start,” so that signal is not sufficient.

This signal belongs to the adjacent case “The report's join stage contains one extreme task that dominates runtime,” whose remedy is Inspect skew metrics and apply AQE/skew mitigation. The observed result for the present requirement is Cold-start component of end-to-end latency decreases.

This signal belongs to the adjacent case “The report reads more files than before despite unchanged filter semantics,” whose remedy is Inspect file pruning, statistics and changed data layout. The observed result for the present requirement is Cold-start component of end-to-end latency decreases.

That observation is a valid indicator for “Repeated recomputation of an intermediate DataFrame dominates a notebook” when using Persist the justified reused intermediate and release after use. The scenario instead tests “The job is spending most time waiting for a cold warehouse to start”; the relevant result is Cold-start component of end-to-end latency decreases.

Inspect skew metrics and apply AQE/skew mitigation belongs to the separate case “The report's join stage contains one extreme task that dominates runtime,” where Straggler duration falls while output counts stay consistent would be expected. That differs from the condition in this stem: “Repeated recomputation of an intermediate DataFrame dominates a notebook.”

Inspect file pruning, statistics and changed data layout is appropriate when “The report reads more files than before despite unchanged filter semantics”; it would be verified by Scan/file metrics identify the additional read volume. It does not directly satisfy “Repeated recomputation of an intermediate DataFrame dominates a notebook,” whose relevant acceptance evidence is Subsequent actions avoid rerunning the expensive ancestor plan.

This action solves a different issue: “A rewrite made the job faster but changes the number of output records.” Its normal signal is The final optimized output agrees with the validated baseline. Here the requested behavior is “Repeated recomputation of an intermediate DataFrame dominates a notebook,” so that signal is not sufficient.

This signal belongs to the adjacent case “A rewrite made the job faster but changes the number of output records,” whose remedy is Run result-equivalence and data-quality regression before accepting the optimization. The observed result for the present requirement is Subsequent actions avoid rerunning the expensive ancestor plan.

This signal belongs to the adjacent case “The job is spending most time waiting for a cold warehouse to start,” whose remedy is Measure startup separately and change supported warehouse readiness settings. The observed result for the present requirement is Subsequent actions avoid rerunning the expensive ancestor plan.

That observation is a valid indicator for “The report reads more files than before despite unchanged filter semantics” when using Inspect file pruning, statistics and changed data layout. The scenario instead tests “Repeated recomputation of an intermediate DataFrame dominates a notebook”; the relevant result is Subsequent actions avoid rerunning the expensive ancestor plan.

Inspect skew metrics and apply AQE/skew mitigation belongs to the separate case “The report's join stage contains one extreme task that dominates runtime,” where Straggler duration falls while output counts stay consistent would be expected. That differs from the condition in this stem: “A rewrite made the job faster but changes the number of output records.”

Inspect file pruning, statistics and changed data layout belongs to the separate case “The report reads more files than before despite unchanged filter semantics,” where Scan/file metrics identify the additional read volume would be expected. That differs from the condition in this stem: “A rewrite made the job faster but changes the number of output records.”

This action solves a different issue: “The job is spending most time waiting for a cold warehouse to start.” Its normal signal is Cold-start component of end-to-end latency decreases. Here the requested behavior is “A rewrite made the job faster but changes the number of output records,” so that signal is not sufficient.

This signal belongs to the adjacent case “The job is spending most time waiting for a cold warehouse to start,” whose remedy is Measure startup separately and change supported warehouse readiness settings. The observed result for the present requirement is The final optimized output agrees with the validated baseline.

Scan/file metrics identify the additional read volume would demonstrate the separate action Inspect file pruning, statistics and changed data layout, which addresses “The report reads more files than before despite unchanged filter semantics.” It does not verify the requirement “A rewrite made the job faster but changes the number of output records” following Run result-equivalence and data-quality regression before accepting the optimization.

This signal belongs to the adjacent case “Repeated recomputation of an intermediate DataFrame dominates a notebook,” whose remedy is Persist the justified reused intermediate and release after use. The observed result for the present requirement is The final optimized output agrees with the validated baseline.