Launch Deal · 50% off 1st month ·

3.1 — Choose a Storage Mode, Including Direct Lake

DP-600 Domain 3 · Skill 3.1

What You Need to Know

Storage mode is a property of each table in a semantic model. It decides where the data lives at query time, how fresh it is, and how fast visuals respond.

The modes

Mode How queries work Freshness Typical use
Import Data is copied into the model (VertiPaq, in memory) on refresh As of the last refresh Most models; any source; full DAX and Power Query features
DirectQuery Each visual sends queries to the source Real time Very large or very volatile data, or a source that must enforce its own security; slower visuals
Dual Behaves as Import or DirectQuery depending on the query Mixed Dimension tables in composite models, so they join efficiently with both Import and DirectQuery facts
Direct Lake VertiPaq loads columns straight from Delta tables in OneLake when queries need them As fresh as the latest framing (seconds) Large Fabric lakehouse or warehouse data with near-Import speed and no data copy

Direct Lake in depth

  • Two flavors:
    • Direct Lake on OneLake can combine tables from several Fabric items, can be mixed with Import tables, and never falls back to DirectQuery.
    • Direct Lake on SQL (analytics endpoint) uses one item, goes through the SQL endpoint, and can fall back to DirectQuery.
  • Framing: a "refresh" of a Direct Lake model only points it at the latest version of each Delta table. That's metadata only and takes seconds. With Keep your Direct Lake data up to date (automatic updates) on, framing happens whenever the tables change. With it off, you reframe with a scheduled or pipeline refresh.
  • Transcoding: columns are loaded into memory on demand, the first time a query needs them, and can be evicted under memory pressure.
  • Fallback (Direct Lake on SQL only) happens for SQL views, warehouse RLS or SQL granular security, or exceeded guardrails (rows, files or row groups per table for the SKU). The DirectLakeBehavior property controls it:
    • Automatic (default): silently falls back.
    • DirectLakeOnly: queries error instead, which is useful in development to find causes.
    • DirectQueryOnly: always DirectQuery, to measure fallback performance.
  • Requirements and limits: a Fabric capacity (F SKU). No Power Query transformations, because data preparation happens upstream. Direct Lake on SQL supports no calculated columns or tables (except calculation groups and field parameters). Direct Lake on OneLake supports some, in preview.
  • Performance depends on the Delta tables: compacted files with V-Order, and sensible row-group sizes.

Choosing

Requirement Choose
Huge gold tables in a lakehouse, near-real-time freshness, no refresh windows Direct Lake
Self-service model mixing several sources with Power Query shaping Import
Must reflect source changes second by second, or data can't leave the source DirectQuery
Import aggregations over a DirectQuery fact, with shared dimensions Composite: DirectQuery fact, Import aggregation table, Dual dimensions (lesson 3.5)

Key Concept: Direct Lake = Import-like speed straight from Delta tables, kept current by framing. Only Direct Lake on SQL can fall back to DirectQuery (views, SQL RLS, guardrails), and DirectLakeBehavior controls that.

Exam Traps

  • Expecting a Direct Lake "refresh" to copy data (it only reframes metadata)
  • Building Direct Lake on SQL views and expecting no fallback
  • Adding Power Query transformations to a Direct Lake table (prepare data upstream instead)
  • Using Import for a multi-terabyte lakehouse table that must be current within minutes

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.

See plans

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