I’ll be following FabConEurope closely (Sept 15–18, Austria Center Vienna) to see how the newest Microsoft Fabric capabilities land in real‑world talks and demos. Here’s the short list I’m watching—centered on near‑zero copy patterns, Materialized Lake Views, Shortcut Transformations, and Real‑Time Intelligence.
Near‑zero copy, done pragmatically
Unify first, then materialize only when it clearly helps. Fabric now gives a solid toolbox to make that practical:
- OneLake shortcuts create a single, virtual lake over data you already store—across ADLS, S3, GCS, and OneLake—so every Fabric engine can read the same files without moving them.
- Cross‑tenant External Data Sharing exposes data in place by creating a shortcut in the recipient tenant (no Entra B2B guest access required). Access is read‑only and governed in the consumer tenant.
- Dataverse ↔ Fabric “Link to Fabric” adds a near real‑time, zero‑copy path from operational apps into OneLake—ideal when you want operational data to show up continuously under Fabric’s governance.
When materialization does buy you something (performance, availability, semantics), there are platform‑native options:
- Mirroring continuously replicates databases (e.g., Snowflake, Azure SQL, Cosmos DB) into OneLake as Delta—analytics‑ready and queryable with SQL/Spark. Open Mirroring extends that with an API/landing zone for change data.
- Warehouse Snapshots (preview) give point‑in‑time, read‑only views you can keep for up to 30 days for stable reporting or reproducible analysis.
- Table Clones in Warehouse provide near‑instant, zero‑copy dev/test clones via T‑SQL.
Materialized Lake Views (MLVs)
What they are. Declarative SQL transforms that Fabric materializes and manages as Delta in the lake—designed to streamline medallion‑style data products with first‑class monitoring. MLVs are in preview.
What stands out right now (per docs):
- Defined with Spark SQL; incremental refresh and API management aren’t available yet (full refresh today).
- You can declare constraints and data‑quality behavior (e.g.,
ON MISMATCH DROP | FAIL) in the create statement. - Runs appear in the Monitor hub with status and execution details; lineage views are supported with noted preview limits.
Why this matters: MLVs push “promote by policy” into definitions you can read and reason about, reducing custom pipeline glue while keeping refreshes and lineage in one place.
Shortcut Transformations (preview)
From “I can see external files” to “I have a governed Delta table”—no bespoke ETL. Shortcut Transformations convert files referenced by a OneLake shortcut (starting with CSV) into managed Delta that stays always in sync. Fabric Spark handles the copy/convert, so you don’t have to orchestrate incremental loads yourself. The service polls the shortcut every ~2 minutes and syncs changes; governance inherits from OneLake (lineage, permissions, Purview policies).
Where it fits: This is the bridge between pure virtualization (shortcuts, shares) and managed, query‑ready tables—useful when you want consistent Delta semantics without spinning up a full pipeline.
Real‑Time Intelligence (RTI): Eventhouse, KQL & real‑time dashboards
What Fabric provides: An end‑to‑end path for events and logs: ingest with Eventstream, store/analyze in Eventhouse/KQL databases, and visualize in Real‑Time Dashboards—all under the same governance model as your lake/warehouse.
Notable recent update: Derived streams → direct ingestion to Eventhouse (preview), configurable from Eventstream or Eventhouse “Get Data,” which simplifies building multi‑stream topologies that land straight into KQL.
Why this matters: It closes the loop—sense → analyze → act—without hopping across services or security models, so real‑time becomes an everyday tool rather than a special project.
One more change to keep in view
Default semantic models are being sunset. Starting Sept 5, 2025, Fabric stopped auto‑creating default semantic models for new warehouses/lakehouses/mirrored items; existing ones convert to regular models you manage explicitly. Plan model creation and governance accordingly.
I’ll be following the keynotes and session recaps for these specific capabilities and any roadmap dates that firm up around them.
