Copyright (c) 2026 MindMesh Academy. All rights reserved. This content is proprietary and may not be reproduced or distributed without permission.
2.1.3. Fabric Items and Their Relationships
💡 First Principle: Fabric items are the building blocks of analytics solutions—like specialized tools in a workshop. Each tool has a specific purpose: you wouldn't use a hammer to measure length, and you wouldn't use a lakehouse for real-time millisecond queries. Understanding which tool to reach for determines whether your solution succeeds or struggles.
Comparative Table: Fabric Item Types
| Item Type | Primary Purpose | Storage Location | Query Language |
|---|---|---|---|
| Lakehouse | Store and process big data | OneLake (Delta tables + Files) | Spark SQL, PySpark |
| Data Warehouse | Structured analytics | OneLake (Delta tables) | T-SQL |
| KQL Database (hosted in an Eventhouse) | Real-time and time-series analytics; a KQL queryset saves KQL queries run against it | OneLake | KQL |
| Eventstream | Real-time data ingestion | Transient (routes to destinations) | Visual editor |
| Data Pipeline | Orchestration | Metadata only | Visual + expressions |
| Dataflow Gen2 | Low-code ETL | OneLake staging | Power Query (M) |
| Notebook | Code-based processing | OneLake (output) | PySpark, Spark SQL |
Visual: Item Relationships in a Typical Architecture
Decision Framework: Which Item to Use When
| If You Need To... | Use This | Why |
|---|---|---|
| Store big data with flexible schema | Lakehouse | Delta tables + raw files, Spark processing |
| Run complex T-SQL analytics | Data Warehouse | Optimized for SQL, familiar to BI teams |
| Analyze streaming data in real-time | KQL Database | Sub-second queries on time-series data |
| Route and transform streaming events | Eventstream | Visual stream processing, multiple outputs |
| Orchestrate multi-step workflows | Data Pipeline | Control flow, scheduling, dependencies |
| Transform data with low-code UI | Dataflow Gen2 | Power Query familiar to Excel users |
| Write custom processing logic | Notebook | Full PySpark/SQL flexibility |
| Run Python ML libraries (scikit-learn, TensorFlow) | Notebook | Spark runtime with Python ML packages |
| Query lakehouse Delta tables with T-SQL, without Spark | SQL analytics endpoint | Created automatically with every lakehouse (read-only T-SQL) |
| Let business analysts build their own models and reports | Semantic model + Power BI report | No engineering skills needed (Power BI datamarts were retired in 2025) |
Common Item Selection Mistakes
| Mistake | Problem | Better Choice |
|---|---|---|
| Using Warehouse for raw file storage | Warehouse requires structured data | Use Lakehouse for files |
| Using Lakehouse for real-time queries | Spark has cold-start latency | Use KQL Database |
| Using Notebook for simple transforms | Overkill, harder to maintain | Use Dataflow Gen2 |
| Using Pipeline when Dataflow suffices | Unnecessary complexity | Dataflow can schedule itself |
⚠️ Exam Trap: Questions often present a scenario and ask which item to use. Key signals: "real-time" → KQL Database/Eventstream; "T-SQL" → Warehouse; "unstructured files" → Lakehouse; "orchestration" → Pipeline.
Enroll to unlock the 7 practice questions written for this section, so you can test what you just read while it is fresh.
Enroll to unlock the 4 flashcards for this section and review them on a spaced-repetition schedule.
Written byAlvin Varughese
Founder•20 professional certifications