Skip to content
DP-700 Fabric Engineer Hub

Exam DP-700: High-Yield Study Cheatsheet

Exam DP-700: High-Yield Study Cheatsheet

Source

  • Provider: Microsoft
  • Platform: Microsoft Learn & Fabric Community
  • Target Exam: Exam DP-700
  • Content type: Quick-Reference Cheatsheet
  • Date captured: 2026-08-15
  • Last reviewed: 2026-08-15

🚀 1. Storage & Table Architecture Quick Matrix

Engine / PatternRead/Write SupportStorage FormatQuery LanguagePrimary Use Case
Fabric Lakehouse (Spark)Read & WriteDelta Lake / ParquetPySpark, Scala, Spark SQLBig Data transformations, ETL/ELT, Data Science
Lakehouse SQL EndpointRead-OnlyDelta Lake / ParquetT-SQLAd-hoc SQL reporting, Power BI Direct Lake
Fabric Data WarehouseRead & WriteDelta Lake / ParquetFull ACID T-SQLEnterprise Star Schema, Cross-Database queries
Eventhouse / KQL DBRead & WriteColumnar CompressedKQL (Kusto)Real-time telemetry, IoT streaming, Log analytics

⚡ 2. PySpark & Delta Optimization Code Snippets

# 1. Read Raw CSV with Schema Enforcement
df = spark.read.format("csv") \
    .option("header", "true") \
    .schema("OrderID INT, OrderDate TIMESTAMP, Amount DOUBLE, Country STRING") \
    .load("Files/raw/*.csv")


# 2. Delta Table Write with Liquid Clustering (or PartitionBy)
df.write.format("delta") \
    .mode("append") \
    .saveAsTable("silver_orders")


# 3. Delta Table Compaction & V-Order
spark.sql("OPTIMIZE silver_orders VORDER")


# 4. Remove Historical Delta Files older than 7 days
spark.sql("VACUUM silver_orders RETAIN 168 HOURS")


# 5. Broadcast Join Optimization
from pyspark.sql.functions import broadcast
joined_df = fact_df.join(broadcast(dim_df), on="Key", how="inner")

📡 3. Real-Time Intelligence & KQL Syntax

// KQL Time-Series Windowing & Anomaly Detection
TelemetryStream
| where Timestamp >= ago(24h)
| make-series AvgLoad = avg(CpuLoad) default=0 on Timestamp from ago(24h) to now() step 15m by Hostname
| extend (Anomalies, Score, Baseline) = series_decompose_anomalies(AvgLoad)
| render anomalychart with(anomalycolumns=Anomalies)

🛡️ 4. Security & Permissions Reference

  • Admin: Full admin control; can manage permissions and delete workspace.
  • Member: Can add contributors/viewers; can edit and execute all items.
  • Contributor: Can create, edit, and execute items; cannot alter workspace roles.
  • Viewer: Read-only access to item data; cannot modify code or run pipelines.
  • Item Permissions:
    • Read: View metadata and query via SQL endpoint.
    • ReadAll: Read all underlying OneLake raw parquet files (bypasses SQL RLS/CLS).
    • Write: Update item definitions and write data.

📊 5. Capacity & Throttling Rules

  • Smoothing Windows:
    • Interactive operations = 5 minutes.
    • Background batch operations = 24 hours.
  • Throttling Escalation:
    1. Interactive Delay (10 min over limit).
    2. Interactive Rejection / HTTP 429 (60 min over limit).
    3. Background Rejection (24 hr sustained overload).

⚖️ 6. High-Yield Pairwise Distinctions

Concept AConcept BKey Exam Distinction
Dataflow Gen2Data PipelineTransformation worker (Power Query) vs. Orchestration manager (triggers/loops/dependencies)
Copy DataDataflow Gen2High-speed data mover without compute overhead vs. visual low-code ETL transformation
LakehouseWarehouseSpark-centric big data/file store vs. relational SQL-first ACID data warehouse
SQL Analytics EndpointWarehouseRead-only Lakehouse SQL query layer vs. full read/write transactional T-SQL
EventstreamEventhouseIn-flight real-time event transformation & routing vs. high-velocity time-series storage
Pipeline ParameterPipeline VariableExternal runtime input vs. internal mutable execution state
Optimize WriteOPTIMIZEWrite-time small file prevention vs. post-load file compaction / bin-packing
V-OrderPartitioningParquet file-level read sorting & indexing vs. directory-level folder hierarchy
Git IntegrationDeployment PipelineSource control synchronization with Git repo vs. multi-stage environment promotion

🔍 7. Warehouse Monitoring DMVs & Query Insights

Diagnostic ObjectScope & Usage
sys.dm_exec_connectionsLive active client connections
sys.dm_exec_sessionsAuthenticated sessions (includes login name, client program)
sys.dm_exec_requestsActively executing T-SQL queries
queryinsights.exec_requests_historyHistorical log of completed queries with CPU/duration metrics
queryinsights.long_running_queriesTop queries ranked by elapsed execution time
queryinsights.frequently_run_queriesTop queries ranked by invocation count

📚 8. Slowly Changing Dimension (SCD) Types

SCD TypeMechanismHistorical Record
Type 0Fixed attribute (e.g. DateOfBirth); updates ignoredInitial value only
Type 1Overwrite existing value in-placeNone (current value only)
Type 2Add new record with StartDate, EndDate, and IsCurrent flagFull historical audit trail
Type 3Add dedicated column for previous value (e.g. PreviousCity)Limited (current + previous only)