DP-700 quick review

Memory hook: Capacity runs the work; workspace organizes it; OneLake stores it.

Must remember

Microsoft Fabric combines analytics experiences around OneLake and shared platform capabilities. A workspace contains items; capacity supplies supported execution resources. A domain organizes business ownership/governance across workspaces rather than acting as an automatic security boundary. Workspace assignment and settings influence where and how jobs run.

A Lakehouse combines files and managed table patterns, commonly Delta, with Spark and supported SQL access. A Warehouse targets relational T-SQL analytical development. Eventhouse/KQL databases target event/time-series analytical workloads. Choose from access pattern, transformation language, transaction needs, latency and operating model, not product name alone.

Configure Spark pools/session behavior, environments/libraries and workspace defaults consistently. OneLake settings affect supported access and integration. Apache Airflow workspace configuration enables its supported orchestration integration; credentials, dependencies and runtime settings still need control. Separate development and production assumptions.

Git integration versions supported item definitions; deployment pipelines promote supported content across stages. They are not automatically a complete backup of every underlying data file or secret. Use database projects for declarative SQL schema development where supported, review differences and manage environment-specific connections/parameters. A successful deployment does not prove target data or permissions are correct.

Choose under exam pressure

Traps

• A workspace is not the same as capacity.
• Git integration does not automatically version every byte of data.

Practise this topic

Memory hook: Workspace access is broad; data rules need their own design.

Must remember

Separate workspace roles, item permissions and data-level controls. A user able to manage an item may have broader access than a consumer intended to see a subset. Apply supported row, column, object and folder/file controls at the layer and engine that actually serves the request. Test every access path, including shortcuts, SQL endpoints, Spark and reports.

OneLake security centralizes supported data-access policies, but feature scope, engine support and configured modes matter. SQL row-level predicates and column/object permissions solve different problems. Dynamic data masking changes displayed values for eligible queries; it is not encryption and not a substitute for restricting privileged/direct access.

Sensitivity labels communicate classification and apply supported protection/handling. Endorsement marks promoted or certified content under organizational governance. Neither replaces permission enforcement or validates business calculations. Audit logs record eligible actions for investigation; configure appropriate collection, retention and access.

Shortcuts reference data without an ordinary duplicate ingestion copy. Access to the shortcut and its target depends on the supported source/credential model. Validate trust boundaries, data residency and external connectivity. A shortcut can fail because its target, connection or permission changed even when its local definition remains intact.

Choose under exam pressure

Traps

• Masked does not mean encrypted.
• One working query path does not prove every engine enforces identical permissions.

Practise this topic

Memory hook: Orchestrate dependencies; move data; make retries safe.

Must remember

Pipelines coordinate activities such as copy, notebooks and dataflows with dependencies, parameters and dynamic expressions. Dataflows Gen2 offer Power Query transformations; notebooks support code-oriented Spark and other supported execution. Choose the tool from complexity, language, scale and maintenance needs rather than assuming one tool fits every step.

Full loads replace or reload a complete dataset; incremental loads process changes since a tracked boundary. A watermark based on a timestamp/key must handle ties, updates, deletes and clock/source semantics. CDC captures changes from a supported source; mirroring replicates supported source changes into Fabric with product-specific limits. Neither guarantees a correct dimensional model automatically.

Use schedules when cadence is time-driven and event triggers when supported events should initiate work. Pass environment and partition parameters explicitly. Model retries, dependencies, concurrency and failure paths; a partially successful run must not duplicate or silently omit records. Stage data, validate it and publish atomically where the engine supports the required semantics.

Track source positions, row counts and business reconciliation totals. Treat late-arriving dimensions/facts deliberately, preserving unknown-member or correction logic. Protect credentials in managed connections and avoid writing secrets into parameters/logs. For Airflow-based orchestration, keep heavy data processing in the appropriate engine rather than passing large datasets through orchestration metadata.

Choose under exam pressure

Traps

• A timestamp watermark can miss records if its boundary logic is wrong.
• A successful copy is not proof that deletes and updates were handled correctly.

Practise this topic

Memory hook: Preserve the grain before optimizing the engine.

Must remember

Raw, cleaned and curated layers often form a bronze/silver/gold pattern. These labels describe responsibilities, not mandatory product features. Delta tables add transactional metadata around supported lake files. Keep schema, partition layout and data quality explicit; simply placing Parquet files in a folder does not automatically create the intended managed Delta table.

Use PySpark for distributed code transformations, T-SQL for supported relational operations and KQL for event/time-oriented analytics. DataFrames are lazily evaluated until an action triggers work. Filter early, select needed columns and avoid collecting a large distributed dataset into a driver process. Joins can produce heavy shuffles or skew around hot keys.

Dimensional models define fact grain, dimensions and surrogate/business keys. Type 1 slowly changing dimensions overwrite the current attribute; Type 2 preserves history with new versions and effective ranges. Late facts need the dimension version valid at event time when historical correctness matters.

Deduplicate using a defensible key and ordering rule, not arbitrary row removal. Distinguish missing values from zero; standardize types/time zones; quarantine invalid records with enough context for repair. Group/aggregate only after deciding what detail can be discarded. Denormalization can improve consumption but may multiply data or complicate updates.

Choose under exam pressure

Traps

• Dropping duplicates without a deterministic rule can keep the wrong version.
• Aggregating early can destroy detail later analysis requires.

Practise this topic

Memory hook: Time boundaries and late events decide the answer.

Must remember

Eventstream ingests, transforms and routes supported real-time events. Eventhouse provides KQL-oriented storage/query capabilities. Spark Structured Streaming offers code-based stateful processing. Select an engine from latency, state, transformation complexity, connectors and team skills.

Native Eventhouse tables ingest data into the engine; OneLake shortcuts reference external lake data. Query acceleration for supported OneLake shortcuts improves suitable query access through additional acceleration behavior and cost/refresh trade-offs. Choose after checking freshness, supported formats and performance needs; it is not identical to ordinary shortcut access.

Event time reflects when an event occurred; processing time reflects when it is handled. Tumbling windows do not overlap, hopping/sliding patterns may overlap, and session windows group activity by inactivity gaps where supported. Watermarks and late-data policies control how long state/results remain open to delayed events. Exact syntax and behavior differ between KQL, Eventstream and Spark.

KQL pipelines use operators such as where, project, extend, summarize and joins to shape event data. Apply selective filters early and choose appropriate time bins. Streaming checkpoints track progress/state for supported recovery, but external side effects still need idempotence. Monitor lag, dropped/late records and poison events rather than only whether the stream is running.

Choose under exam pressure

Traps

• A running stream can be far behind real time.
• A window result can be incomplete if late data is discarded.

Practise this topic

Memory hook: Locate the failing layer, then optimize the measured bottleneck.

Must remember

Monitor ingestion, transformation, semantic-model refresh and consumption. Track job state, duration, freshness, completeness, rejected records and capacity utilization. Alerts need owners and actionable thresholds. Capacity contention can slow several unrelated items at once, while a single malformed record may fail only one pipeline.

For pipeline/Dataflow failures, inspect the failing activity/step, parameters, credentials, gateway/connection, schema and source throttling. For notebooks, examine execution logs, library/environment versions, resource exhaustion and data paths. Eventstream/Eventhouse failures may involve source connections, schemas, transformation expressions or ingestion policies. T-SQL errors need actual statement/schema/permission evidence; shortcut errors need target/connection checks.

Optimize Lakehouse tables by addressing small files, partition design and supported compaction/layout options. V-Order targets suitable columnar read efficiency; it is not a universal remedy for every write-heavy workload. Cleanup such as VACUUM affects historical-file availability, so retention and readers matter. Do not delete underlying files manually to solve a table error.

Tune Spark partitions, join strategies, skew, caching and executor resources from observed stages. Tune warehouse queries by reducing unnecessary scans and expensive joins and checking concurrency. For streaming, examine throughput, batch/window/state behavior and backpressure. Pipelines benefit from appropriate parallelism, not unlimited simultaneous source requests.

Diagnosis drill: a daily dashboard is stale despite a green copy activity. Check transformation completion, target partitions, semantic-model refresh and report access in order. End-to-end correctness crosses more than one successful task.

Choose under exam pressure

Traps

• More parallelism can overwhelm a source or capacity.
• A green upstream task does not prove a fresh downstream report.

Practise this topic