5.2 High-Yield Traps & Confusions
Key Takeaways
- Direct Lake vs Import: Direct Lake reads OneLake files with no import copy and stays fresh via framing; Import caches data and needs scheduled refresh.
- RLS filters rows, CLS hides columns, OLS hides whole tables/objects — the exam pairs the access need to the right mechanism.
- Dataflow Gen2 is low-code Power Query transformation; a pipeline orchestrates and moves data — they solve different problems and are often combined.
- Promoted is a lightweight 'recommended' signal any contributor can set; Certified is a governed approval restricted by admins.
- Incremental refresh applies to Import tables, not Direct Lake; matching the technique to the storage mode avoids a classic trap.
Where DP-600 Sets Its Traps
Microsoft writes plausible distractors from closely related Fabric features. If you can state the precise difference in one line, multi-choice and case-study questions become much easier.
The High-Yield Confusion Table
| Pair | The deciding rule |
|---|---|
| Direct Lake vs Import | Direct Lake reads OneLake Delta files in place, no copy, fresh via framing; Import caches a compressed copy and needs scheduled refresh |
| Direct Lake vs DirectQuery | Direct Lake is the in-memory fast path; on a query it cannot serve, it falls back to DirectQuery against the SQL endpoint |
| RLS vs CLS vs OLS | RLS filters rows; CLS hides specific columns; OLS hides whole tables/objects from a role |
| Dataflow Gen2 vs Pipeline | Dataflow Gen2 = low-code Power Query transform; Pipeline = orchestration/movement of activities |
| Lakehouse vs Warehouse vs Eventhouse | Lakehouse = open Delta + Spark; Warehouse = T-SQL relational; Eventhouse = real-time KQL telemetry |
| Promoted vs Certified | Promoted = lightweight recommendation any contributor sets; Certified = governed approval limited to authorized reviewers |
| Sensitivity label vs permission | A label classifies/protects information; a permission grants or denies access — a label is not access control |
| Calculation group vs Field parameter | Calc group changes how a measure is calculated; field parameter changes which field/measure is shown |
Storage Mode Matched to Refresh Technique
A recurring trap pairs the wrong maintenance technique with a storage mode. Incremental refresh partitions and refreshes Import tables — it is not how Direct Lake stays current. Direct Lake freshness is governed by framing/reframe. Likewise, a DirectQuery model has nothing to import-refresh; it is always live. When a question proposes 'configure incremental refresh on the Direct Lake model,' that is almost always the distractor.
Endorsement Is Not Deployment, Classification Is Not Security
Two governance confusions appear often. Endorsement (Promoted/Certified) signals trust in content; it does not move content between environments — deployment pipelines do that. A sensitivity label classifies and can encrypt content for compliance; it does not grant or restrict who can open an item — workspace roles and item permissions do. Keeping 'trust signal vs promotion mechanism' and 'classification vs access control' separate clears a cluster of questions.
A governance lead wants only an authorized review team to be able to mark a semantic model as the official enterprise source, while still letting any contributor flag content as recommended. Which endorsement design fits?
A report must hide salary figures (a column) from one group and remove an entire confidential 'Audit' table from another group's view. Which controls apply, respectively?