DP-700 Course Notes

Spark pool questions usually ask which setting controls compute shape, concurrency, or elasticity. Custom Spark pools let a workspace use known node sizes for heavier notebook workloads, while workspace default Spark pools keep routine jobs from accidentally landing on oversized or undersized compute. Autoscaling is the clue when demand changes across runs and the pool should add or remove nodes instead of being manually resized each time.

No common mistakes are available for this section yet.

No exam tips are available for this section yet.

Native Fabric deployment pipelines move workspace items through development, test, and production stages. Use them when the same item set must be promoted with controlled stage permissions and environment-specific settings. The exam distinction is that a deployment pipeline is the Fabric release mechanism, while Git integration tracks source-controlled changes before those items are deployed.

No common mistakes are available for this section yet.

No exam tips are available for this section yet.

Sensitivity labels can flow downstream from a semantic model, but label inheritance does not mean every downstream item is permanently locked to that label. DP-700 questions usually test whether a downstream label should inherit, be overridden, or become more restrictive. The safe answer depends on the label rule and whether the downstream item contains data that requires stronger classification.

No common mistakes are available for this section yet.

No exam tips are available for this section yet.

Fabric performance depends on whether a workspace is backed by shared user licensing or dedicated capacity. Pro licensing is tied to user access, while Premium or F-SKU capacity gives workloads dedicated Fabric capacity resources. If a question links slow execution to workspace placement, licensing, or capacity assignment, the answer usually starts with the capacity backing the workspace.

No common mistakes are available for this section yet.

No exam tips are available for this section yet.

Medallion architecture separates raw, cleaned, and curated data into Bronze, Silver, and Gold layers. Bronze keeps source-shaped data for traceability, Silver applies cleansing and standardization, and Gold is shaped for reporting or downstream consumption. The exam clue is the amount of transformation: land first, clean next, serve last.

No common mistakes are available for this section yet.

No exam tips are available for this section yet.

Spark Structured Streaming can read from Fabric Eventstreams and write events into a Lakehouse destination backed by Delta tables. Use this pattern when the data engineering job needs Spark transformations over streaming data rather than only routing events through the Eventstream experience.

No common mistakes are available for this section yet.

No exam tips are available for this section yet.

A semantic model is the analytical layer that organizes tables, relationships, measures, and hierarchies for reporting. In DP-700, the model structure question usually asks how data should be navigated, while the security question asks which users should see which rows or entities.

No common mistakes are available for this section yet.

No exam tips are available for this section yet.

Eventstream routing decides where real-time events go after ingestion. A stream can route telemetry to a KQL database for low-latency analysis, to a Lakehouse for storage, or through transformation operators before landing. The exam clue is whether the requirement is routing, transforming, storing, or querying the stream.

No common mistakes are available for this section yet.

No exam tips are available for this section yet.

Monitor Hub is the first place to check when a Fabric pipeline or Copy Activity fails. It shows run status, timing, and activity details so the error can be tied to the failing activity instead of guessed from the whole pipeline. For copy failures, inspect the source, sink, authentication, schema, and row-level error details shown in the run output.

No common mistakes are available for this section yet.

No exam tips are available for this section yet.

Warehouse tuning starts with the visual execution plan when a T-SQL query is slow. The plan shows joins, scans, movement, and expensive operators so the fix can target the real bottleneck. Statistics and column statistics help the optimizer estimate row counts; stale or missing statistics can lead to poor join choices even when the SQL text looks reasonable.

No common mistakes are available for this section yet.

No exam tips are available for this section yet.