DP-750 Practice Test

Sample DP-750 Practice Questions

10 real questions from the free DP-750bank. Pick an answer in your head, then open “Show answer” for the explanation.

1. 1. An organization wants to configure cluster autoscaling to balance cost efficiency with the ability to absorb workload spikes on mission-critical pipelines. Which configuration reflects documented Well-Architected guidance for Azure Databricks?

   • A. A minimum node count of zero with no maximum, so the cluster can scale freely in either direction
   • B. A minimum node count equal to the maximum node count to guarantee consistent performance
   • C. A minimum node count of two, a maximum aligned to workspace quota, and target utilization between 70 and 80 percent
   • D. A fixed node count equal to the maximum expected load, with autoscaling disabled
   Show answer

   Correct answer: C

   Well-Architected guidance for Azure Databricks recommends setting a minimum node count of two and a maximum node count aligned with workspace quota limits, with target utilization thresholds between 70% and 80% to balance cost efficiency and performance headroom. A minimum of zero risks slow cold starts for critical pipelines. Setting minimum equal to maximum effectively disables autoscaling. A fixed node count with autoscaling disabled removes the cost benefits of dynamic scaling.

2. 2. A team runs DROP TABLE on an external table registered in Unity Catalog. What happens to the underlying data files in cloud storage?

   • A. The data files are permanently deleted along with the table metadata
   • B. Only the table's metadata registration is removed; the underlying data files in cloud storage remain
   • C. The data files are moved to a metastore-level trash location for 30 days before deletion
   • D. The operation fails unless the external location is first unregistered
   Show answer

   Correct answer: B

   Dropping an external table removes only its metadata registration from Unity Catalog; because the data itself lives at a cloud storage path outside of Unity Catalog's management, the underlying files are left in place. This is the key distinction from a managed table, where dropping the table also removes the underlying data. Option A describes managed-table behavior. Option C invents a trash-retention mechanism that isn't part of dropping an external table. Option D is false - dropping the table registration does not require unregistering the broader external location first.

3. 3. A team creates a brand-new Azure Databricks workspace with Unity Catalog enabled. Without creating anything manually, what do they typically find already available in Catalog Explorer?

   • A. A workspace catalog automatically provisioned and named after the workspace
   • B. Nothing; catalogs must be created manually before any data can be stored
   • C. A sample reference catalog containing Databricks tutorial datasets
   • D. A foreign catalog pre-connected to the organization's primary data warehouse
   Show answer

   Correct answer: A

   New workspaces are automatically provisioned with a workspace catalog, by default named after the workspace itself, which teams can use immediately without creating additional catalogs right away. Option B is incorrect since this default catalog exists without any manual action. Option C misdescribes the catalog's purpose -- it is a general-purpose working catalog, not a tutorial dataset reference. Option D is false because no foreign catalog connection is automatically created for any organization.

4. 4. A team defines a materialized view using CREATE OR REFRESH MATERIALIZED VIEW that reads from a Unity Catalog managed Delta table with row-tracking enabled, using only SELECT, WHERE, and GROUP BY. What refresh behavior can this materialized view support?

   • A. A full refresh; incremental refresh is not available to materialized views
   • B. Incremental refresh, since Delta tables with the required operations and row-tracking are eligible sources for incremental refresh
   • C. Incremental refresh if the source table is a legacy Hive metastore table
   • D. Incremental refresh if the materialized view uses a recursive common table expression
   Show answer

   Correct answer: B

   Delta tables, including Unity Catalog managed tables, are supported data sources for incremental refresh, and SELECT, WHERE, and GROUP BY are among the supported clauses, with row-tracking enabled where required, so this materialized view is eligible for incremental refresh rather than requiring a full recompute. Option A is factually wrong, since incremental refresh is explicitly supported for eligible materialized views. Option C reverses the actual requirement -- Hive metastore tables are not the basis for incremental refresh eligibility. Option D is incorrect because materialized views using recursive CTEs are specifically NOT eligible for incremental refresh.

5. 5. A data scientist tries to enable both the Photon acceleration checkbox and a GPU-enabled worker type on the same compute resource. What happens?

   • A. Both settings are applied together, and Photon accelerates the GPU workload as expected
   • B. The GPU worker type is ignored, and the compute resource provisions CPU instances instead
   • C. This combination is unsupported; Photon acceleration is not supported with GPU instance types
   • D. Photon silently falls back to CPU execution while GPU processing continues normally
   Show answer

   Correct answer: C

   Photon acceleration is not supported with GPU instance types, so when configuring GPU compute, the Photon acceleration checkbox must be unchecked. Option A describes a combination that is explicitly unsupported, not an expected successful outcome. Option B incorrectly claims the GPU worker type gets ignored rather than the Photon setting being the incompatible element. Option D invents a silent fallback behavior that isn't documented.