2.3 — Choose a Store, Use OneLake Integration and Partition Files
DP-600 Domain 2 · Skill 2.3
What You Need to Know
An analytics engineer chooses where curated data lives (lakehouse, warehouse or eventhouse), knows how to make data from any engine available in OneLake, and lays out large lakehouse tables so that queries read less.
Lakehouse, warehouse or eventhouse?
| Choose | When |
|---|---|
| Lakehouse | Files of any shape plus Delta tables; Spark-first teams; data science; a medallion architecture. T-SQL users read it through the read-only SQL analytics endpoint |
| Warehouse | SQL-first teams; full T-SQL DML, stored procedures and multi-table transactions; a star schema for BI |
| Eventhouse | Streaming, telemetry, logs and time-series data that needs analysis within seconds with KQL |
All three can feed Direct Lake semantic models (an eventhouse once OneLake availability is on).
OneLake integration
| Feature | What it does |
|---|---|
| Eventhouse OneLake availability | Writes KQL tables to OneLake as Delta (in batches, up to about 3 hours by default), so lakehouses, warehouses, notebooks and Direct Lake can use them. Choose Apply to existing tables to backfill |
| Semantic model OneLake integration | An Import model writes its tables to OneLake as Delta tables after each refresh. Turn it on in the model settings (F or P capacity only). Measures, DirectQuery tables and calculation groups aren't exported |
| Shortcuts to those Delta tables | Let a lakehouse or warehouse query them without copying |
A common use: a Power Query-heavy Import model already cleans data well. With OneLake integration, data engineers can reuse those cleaned tables in notebooks and SQL.
File partitioning in a lakehouse
(sales_df.write.format("delta")
.partitionBy("Year", "Month")
.mode("overwrite")
.saveAsTable("sales"))
- Partitioning writes one folder per value (
Year=2026/Month=11/). Queries that filter on the partition columns read only those folders (partition pruning). - Partition only large tables, by low-cardinality columns that queries filter on, such as date parts or region. A guideline is at least about 1 GB of data per partition.
- Don't partition by high-cardinality columns (CustomerID, a timestamp). That creates thousands of tiny files, which slows Spark, SQL and Direct Lake.
- The same idea applies to raw files under Files: organize them as
Files/raw/sales/2026/11/09/...so that loads and queries read only the folders they need.
Key Concept: Choose the store from the team's language and the workload. OneLake integration lets any engine's data become Delta in OneLake. Partition only big tables, by low-cardinality filter columns.
Exam Traps
- Choosing a lakehouse when stored procedures must update several tables in one transaction
- Partitioning a 5-million-row table by CustomerID
- Expecting OneLake integration to export measures or DirectQuery tables
- Expecting eventhouse data in OneLake immediately (writes are batched)
Check your understanding
Practice Scenario
🔒 This practice scenario is part of Skillmarked Pro
The lessons stay free. Pro unlocks the other 17 DP-600 practice scenarios and the 50-question timed mock exam, plus everything for every other exam.
Launch Deal: first month €9.95, then €19.90/month. Enter code LAUNCH50 at checkout. Ends 1 December.
Cancel anytime. Or €199 a year, which saves 17%.
Not sure yet? Try 10 free mock exam questions.
Already Pro? Log in.
Sources
- Fabric decision guide: choose a data store, Microsoft Learn
- OneLake integration for semantic models, Microsoft Learn
- Turn on OneLake availability for an eventhouse, Microsoft Learn
- Load data to lakehouse using partition, Microsoft Learn