DP-700 Skills Measured and Fabric Workflows

Fabric work starts in workspaces. Candidates should know how workspaces group lakehouses, warehouses, pipelines, notebooks, semantic models, and related items. Domains can help organize data estates at a broader level. A common mistake is treating a workspace as only a folder instead of a security, collaboration, and lifecycle boundary.

OneLake is the shared data foundation for Fabric. Lakehouses use files and Delta tables for engineering workflows, while warehouses provide relational analytics experiences. The decision depends on access pattern, transformation approach, SQL needs, and how downstream consumers will query or reuse the data.

Lakehouse work includes loading data, creating Delta tables, transforming data with Spark, managing schema, and optimizing tables. Z-Order can improve data skipping for selected query patterns when used appropriately. Candidates should connect table design and optimization to the queries users actually run.

Warehouses support T-SQL analytics, relational modeling, and structured reporting needs. Materialized views can help when precomputed results improve repeated analytical queries. A DP-700 scenario may test when a warehouse pattern fits better than notebook-heavy engineering or file-oriented lakehouse processing.

Data Factory pipelines orchestrate activities, copy data, and coordinate multi-step movement. Dataflows Gen2 support low-code transformation and shaping. The right tool depends on who maintains the workflow, how complex the transformations are, and whether orchestration or transformation is the main requirement.

Spark notebooks support code-first data preparation, exploration, and engineering. Candidates should understand when notebooks are useful for transformations, testing logic, and building repeatable steps. Notebook workflows should still fit governed deployment and monitoring practices instead of remaining one-off experiments.

Real-time scenarios require different tools than scheduled batch processing. Eventstreams help ingest streaming events, while Real-Time Intelligence supports analysis over event-driven data. Candidates should recognize when the requirement is continuous event handling rather than periodic pipeline refresh.

Mirroring can make source data available in Fabric for analytics with less manual pipeline construction. The decision depends on source support, freshness needs, and how the data will be consumed. Candidates should avoid choosing mirroring when the requirement is custom transformation logic or complex orchestration.

Fabric engineering includes permissions, workspace roles, item access, Git integration, and deployment pipelines. Git supports versioned collaboration. Deployment pipelines support movement across stages. Security and release flow matter because data products usually move from development to shared production use.

Monitoring includes pipeline runs, refresh behavior, performance symptoms, and capacity usage. The Fabric Capacity Metrics app helps teams inspect capacity behavior. Candidates should know how to reason about slow queries, failed runs, throttling symptoms, and whether the fix belongs in table optimization, orchestration, query design, or capacity planning.