Free DP-700 Exam Flashcards

A workspace is the collaboration and security boundary where Fabric items such as lakehouses, warehouses, notebooks, pipelines, and semantic models are created and managed.

A workspace role controls what a user can do with items in that workspace. Use the least privileged role that supports the user's job task.

A deployment pipeline promotes Fabric item definitions through lifecycle stages such as development, test, and production. It moves metadata and configuration, while environment-specific data and credentials still need separate handling.

A deployment rule changes stage-specific settings, such as connections or parameters, so promoted items point to the correct environment after deployment.

Git integration lets teams version supported Fabric workspace items, review changes, and use branch-based development practices before deployment.

OneLake is Fabric's tenant-wide data lake. It gives Fabric items a shared storage foundation so data can be reused without unnecessary copies.

A lakehouse stores files and Delta tables for analytics. It is the common Fabric choice when engineers need open file storage plus table-oriented processing.

The Files area holds raw or semi-structured files, while the Tables area exposes managed Delta tables for query and analytics workloads.

A shortcut references data stored in another location without copying it into the lakehouse. Use shortcuts to simplify access while respecting the source location's security and performance constraints.

A Delta table combines data files with a transaction log. The log supports reliable table operations, schema handling, and consistent reads for analytics workloads.

A Fabric warehouse is a SQL analytics store for relational modeling, T-SQL development, and serving curated data to reporting and downstream consumers.

The SQL analytics endpoint lets users query lakehouse tables with SQL. It is useful for read-oriented exploration and reporting over lakehouse data.

The grain defines what one fact row represents. Clear grain prevents double counting and keeps measures consistent across dimensions.

A dimension table stores descriptive attributes such as customer, product, date, or region. Dimensions make facts easier to filter, group, and understand.

A stored procedure can package repeatable SQL transformation logic for a pipeline or warehouse workflow. It is useful when the logic belongs close to relational tables.

A full load replaces or reloads the whole target dataset. It is simple, but it can waste time and capacity when only a small portion of the source changes.

An incremental load moves only new or changed data since the last successful run. It reduces processing work when the source provides a reliable change indicator.

A watermark records the point up to which source data has already been processed. Pipelines and transformations use it to resume without reloading everything.

Late-arriving data reaches the pipeline after related records or reporting windows have already been processed. A resilient design loads it safely and updates affected tables instead of silently dropping it.

Mirroring brings supported operational source data into Fabric for analytics with less custom ingestion code. It is best when the goal is ongoing replication rather than one-time file movement.

A pipeline orchestrates data movement and processing activities. It coordinates tasks such as copy operations, notebooks, dataflows, stored procedures, and dependency flow.

Copy activity moves data between supported sources and destinations. It is a practical choice when the main task is data movement rather than complex transformation.

A pipeline parameter supplies values at run time, such as dates, folder paths, or environment names. Parameters make one pipeline reusable across runs and stages.

Activity dependencies define the order and conditions for pipeline execution. They help separate success paths, failure handling, and cleanup work.

A notebook activity runs Spark code as part of an orchestrated workflow. Use it when transformations need PySpark, Scala, or notebook-based engineering logic.

A Fabric notebook is an interactive development item for Spark code, SQL, markdown, and exploratory engineering. It can be run manually or orchestrated in a pipeline.

A Spark session is the execution context for notebook commands. Session startup, configuration, and reuse affect both developer experience and workload cost.

A PySpark DataFrame represents distributed tabular data. It is used for scalable transformations that are too large or complex for simple row-by-row processing.

Partitioning controls how data is divided for distributed processing. Good partition choices improve parallelism, while poor choices can cause skew or unnecessary shuffling.

A checkpoint stores streaming progress and state. It lets a streaming job resume after interruption without losing its place or blindly starting over.

Dataflow Gen2 uses Power Query to prepare and transform data in Fabric. It is a strong fit for low-code shaping, cleansing, and loading tasks.

Power Query applies steps in sequence. Changing an early step can affect later filters, joins, data types, and destination mapping.

Append stacks rows from compatible tables. Merge joins related tables by key to add columns from another dataset.

A dataflow destination defines where transformed results are loaded, such as a lakehouse table or warehouse table. The destination choice should match how the data will be queried and governed.

Schema drift happens when source columns or types change unexpectedly. Dataflows should handle expected changes deliberately so refresh failures do not hide data-quality problems.

An eventstream ingests and routes event data in Fabric. It is used when records arrive continuously and need real-time processing or storage.

An eventhouse stores and serves high-volume event and telemetry data for KQL querying. It is suited to time-oriented operational analytics.

KQL is a query language for exploring event, log, and telemetry data. DP-700 candidates should recognize when KQL is the right tool instead of T-SQL or PySpark.

A windowed aggregation groups streaming events by time window. It supports metrics such as counts, averages, and anomaly signals over recent activity.

A streaming destination is where processed events land, such as an eventhouse, lakehouse, or other supported Fabric target. Choose it based on query pattern, latency needs, and retention requirements.

Least privilege means granting only the access needed for a task. In Fabric, this applies across workspace roles, item permissions, data access, and source credentials.

OneLake security controls data access within the lake. It matters because seeing a Fabric item does not always mean a user should read every underlying file or table.

A sensitivity label identifies protected or regulated data. Labels help users and governance tools understand handling requirements for Fabric items.

Endorsement marks trusted content for reuse. Promoted content signals owner confidence, while certified content usually reflects stronger organizational review.

A domain organizes data and analytics content by business area. It improves discovery and ownership clarity without automatically granting access.

The monitoring hub helps track runs, refreshes, and workload activity across Fabric items. Use it to find failures, long-running jobs, and operational trends.

Pipeline run history shows activity status, timing, and errors. It is the first place to check when an orchestration fails or misses its expected completion window.

Query tuning reduces unnecessary scans, joins, shuffles, and transformations. In Fabric, the right fix depends on whether the workload is SQL, Spark, KQL, or semantic-model driven.

Too many small files can slow lakehouse reads because engines spend extra work opening and planning files. Compaction and better write patterns can improve scan efficiency.

Fabric workloads share capacity resources. Efficient schedules, right-sized compute, and workload monitoring help prevent one engineering job from disrupting other analytics users.