Skip to content

DP-700 Fabric Data Engineer Associate - Fact Sheet

Quick Reference

Exam Code: DP-700 Duration: 100 minutes Format: Multiple choice + multiple response + case studies + Labs (some sittings) Cost: $165 USD Passing Score: 700/1000 Validity: 1 year (free annual renewal via Microsoft Learn)

📖 Official DP-700 page 📖 Skills Measured 📖 Microsoft Fabric Documentation

Microsoft Fabric architecture

Fabric unifies several Azure analytics services into a SaaS platform with shared storage (OneLake), unified billing (capacity units), and a single experience for engineering, science, BI, and real-time analytics.

Workloads

Workload What it is
Data Engineering Lakehouse, notebooks (Spark), Spark job definitions
Data Factory Pipelines, Dataflows Gen2 (Power Query)
Data Warehouse T-SQL warehouse with auto-distributed compute
Real-Time Intelligence Eventstreams, KQL databases (Kusto), real-time dashboards
Power BI Reports, dashboards, semantic models
Data Science ML notebooks, models, experiments
Databases (preview) OLTP databases (SQL DB) inside Fabric

OneLake

  • Single, organization-wide data lake on top of ADLS Gen2
  • "OneDrive for data"
  • Default format: Delta Parquet
  • Items in workspaces (Lakehouses, Warehouses) all sit in OneLake
  • Shortcuts point to data in other Fabric workspaces or external systems (ADLS, S3, GCS, Dataverse)

Capacity

  • Compute is purchased as Capacity Units (CU) at SKU sizes (F2, F4, F8, F16, F32, F64, ...)
  • Trial: F64 capacity for 60 days free
  • Compute is shared across all workloads in workspaces assigned to a capacity
  • Throttling kicks in when burstable consumption exceeds capacity for too long

Ingestion patterns

Data Pipelines (Data Factory)

  • Visual orchestration; copy activity, dataflow activity, notebook activity, lookup, foreach, etc.
  • 100+ connectors (similar to ADF)
  • Schedule-based, event-based, or on-demand triggers

Dataflows Gen2

  • Power Query-based, M language
  • Low-code; great for analyst/data-engineer-lite use
  • Output to Lakehouse, Warehouse, KQL DB, or Azure SQL DB

Notebooks (Spark)

  • PySpark, Spark SQL, R, Scala
  • Read/write Lakehouse Delta tables
  • Schedule via pipelines or Spark Job Definitions

Eventstream

  • Streaming ingestion: Event Hubs, IoT Hubs, Kafka, Azure SQL CDC, AWS Kinesis, Google Pub/Sub
  • Routes to Lakehouse, KQL DB, or Eventhouse
  • Built-in transformations (filter, aggregate, manage fields)

Mirroring

  • Continuous replication of operational data into Fabric (zero-ETL)
  • Sources: Azure Cosmos DB, Snowflake, Azure SQL DB
  • Lands as Delta tables in OneLake; queryable immediately

Lakehouse vs Warehouse

Feature Lakehouse Warehouse
Storage Files + Delta tables Delta tables only
Engine Spark + SQL endpoint (read-only) Native T-SQL (read+write)
Use case Engineering, ML, semi-structured Analytics, BI, structured
DML support Spark writes; SQL endpoint is read-only Full T-SQL DDL/DML
Schema Delta schema, schema-on-read for files Strict T-SQL schema
Compute Spark MPP T-SQL engine
Cost model Pay capacity for compute Same

Choose Lakehouse for raw + curated zones, Spark / Python development, ML. Choose Warehouse for strict T-SQL, BI-style aggregations, transactional workloads inside the analytics layer.

Both share OneLake, so Power BI Direct Lake mode reads either.

Security and governance

Workspace roles

Role Permissions
Admin Full control
Member Edit content, manage sources
Contributor Edit content
Viewer Read-only

Item permissions

Beyond workspace roles, individual items (Lakehouse, Warehouse, Pipeline) can have explicit Read / ReadAll / Write / Reshare permissions.

OneLake security (preview)

Folder/table-level access controls using OneLake security predicates for row/column filters.

Sensitivity labels

Microsoft Purview Information Protection labels (Public, Confidential, Highly Confidential, etc.) propagate from sources into Fabric content for downstream policy enforcement.

Microsoft Purview integration

  • Data Catalog auto-discovers Fabric items
  • Lineage tracks item-to-item flow
  • Data quality and policy

Deployment Pipelines (CI/CD)

  • Promote items between Dev → Test → Prod workspaces
  • Compare changes before deployment
  • Rules to swap parameters per environment (data source, capacity)
  • Git integration: backing your workspace with a repository (Azure DevOps or GitHub) for full version control

Monitoring

Monitoring Hub

  • Centralized view of all activities (pipeline runs, Spark jobs, dataflows, refreshes)
  • Filter by workspace, item, status, time range
  • Drill into individual run logs

Capacity Metrics App

  • Free Power BI app showing per-capacity CU consumption
  • Identifies which items / users / time windows consumed the most
  • Throttling indicators

Common exam triggers

  • "Continuous CDC from Cosmos DB into Fabric without code" → Mirroring
  • "Streaming events from IoT Hub" → Eventstream
  • "Low-code Power Query transform" → Dataflows Gen2
  • "PySpark notebook on Lakehouse data" → Notebooks
  • "Strict T-SQL writes" → Warehouse, not Lakehouse SQL endpoint
  • "Promote workspace from Dev to Prod" → Deployment Pipelines
  • "Reference S3 data without copying" → Shortcut to S3
  • "Multi-workspace shared semantic model" → Power BI semantic model + Direct Lake
  • "Restrict columns in a Lakehouse table" → OneLake security predicates / row-column filters
  • "Schedule a notebook nightly" → Pipeline with Notebook activity, scheduled trigger

Highest-yield facts

  1. OneLake is the single Fabric storage layer. Every Lakehouse, Warehouse, KQL DB lives in it as Delta files.
  2. Direct Lake is Power BI's Fabric-native storage mode - no import, no DirectQuery, reads Delta directly from OneLake. Fastest-perf BI when everything is in Fabric.
  3. Capacity throttling: Fabric has burstable + smoothing windows. Sustained over-consumption = throttle. Right-size capacity SKU.
  4. Lakehouse SQL endpoint is read-only; for T-SQL writes use a Warehouse.
  5. Mirroring is zero-ETL - no pipeline, no Spark, no copy activity. Just a config + landing in OneLake.
  6. Shortcuts to external storage = no copy, queries fan out. Useful for federated queries.
  7. Deployment Pipelines support Dev/Test/Prod, item-level diff, and rule-based parameter swaps.
  8. Git integration = workspace ↔ repo sync; full version control for Fabric items (notebooks, pipelines, semantic models).
  9. Eventstream has built-in transforms (filter, aggregate, manage fields) - you don't need a separate stream processor for simple transforms.
  10. OneLake security is the row/column-filter mechanism (preview as of 2026).

Comparison: DP-700 vs DP-203

Aspect DP-700 (Fabric) DP-203 (Azure)
Platform Microsoft Fabric (SaaS) Synapse, ADLS Gen2, ADF (Azure-native)
Storage OneLake (single lake) ADLS Gen2 (per-account)
Engine Spark + Warehouse + KQL Synapse Spark + SQL Pool + Synapse Pipelines
Power BI Native (Direct Lake) Imported / DirectQuery
Audience Data engineers using Fabric SaaS Data engineers using Azure-native services
Status Modern, growing Still active; companion path

Most candidates today take DP-700 if their org is on Fabric; DP-203 if on classic Azure data stack.

Hands-on practice priorities

  1. Build a Lakehouse: ingest CSVs to Files, convert to Delta tables via notebook
  2. Create a Pipeline: schedule a notebook + Dataflow + email-on-failure
  3. Build a Dataflow Gen2: Power Query-style transform writing to Warehouse
  4. Set up Mirroring from a free Cosmos DB account
  5. Configure Eventstream from sample event source → Lakehouse + KQL DB
  6. Promote a workspace through a Deployment Pipeline (Dev → Test → Prod)
  7. Connect the workspace to a Git repo and commit changes
  8. Build a Power BI report on a Lakehouse with Direct Lake mode

These cover ~80% of likely exam scenarios.

Comparable certs in this repo