OneLake Shortcuts: Data Mesh Without the Copy-Paste

The Executive Insight

Most organizations built their first data mesh the expensive way. Every new domain meant another copy of the data, another storage account, another pipeline, and another permission set to babysit. Decentralized ownership turned into centralized maintenance.

OneLake shortcuts are Fabric’s quiet correction to that. They let a domain reference data in another lakehouse, warehouse, workspace, or an external store like ADLS Gen2 or S3 — without copying it — while everything still presents as a single virtual lake.

Issue #18 is about what that changes architecturally: how to get domain-oriented analytics without standing up a full-time data-logistics function to keep duplicates in sync.

What a Shortcut Actually Is (And Isn’t)

A shortcut is a governed pointer — effectively a symbolic link inside your lakehouse — to data that lives somewhere else: another Fabric item, an Eventhouse, or an external Delta/Iceberg store in ADLS Gen2 or S3.

The properties that matter for an architecture decision: no duplication by default, so primary control stays with the source; a unified namespace, so consumers don’t need to know where the physical store lives; table awareness for Delta and Iceberg; and security that respects the source rather than bypassing it. Internal Fabric shortcuts use the caller’s identity; external ones use the connection’s credentials.

The one-line version for a steering committee: a shortcut is not another copy of the data — it’s a governed pointer that lets multiple domains treat one physical dataset as local while the owner stays in charge.

Why This Matters for Data Mesh

“Single virtual data lake” is the marketing phrase. Shortcuts are the mechanism that makes it real, and the implication for a mid-market estate is specific:

Domain teams keep their own workspaces, lakehouses, and semantic models, but plug into shared upstream data via shortcuts instead of cloning it. A central gold lakehouse becomes the canonical source — published once, consumed everywhere. And existing ADLS or S3 data joins without a migration project.

The result is the collaboration and discovery benefits of a mesh without funding a multi-year migration or a standing fleet of copy jobs. You can behave like you have one well-organized lake even while some data sits in ADLS, some in S3, and some in other workspaces — and you’re not ready to move any of it yet.

How Shortcuts Change Architecture Decisions

Once shortcuts are on the table, several default moves start to look like legacy habits. You no longer copy data into every consuming workspace, maintain a separate physical gold per team, or migrate an external lake on day one just to use Fabric tooling.

The patterns shift accordingly:

• From “copy into my workspace” → “shortcut to the owner’s lakehouse.” A sales analytics lakehouse points at central customer and product gold tables rather than reproducing them.
• From “rebuild pipelines for external lakes” → “shortcut to ADLS Gen2 / S3.” Trusted external tables surface into Fabric without rewriting ingestion.
• From “multiple golds” → “one gold, many views.” A single canonical store feeds many domain lakehouses, each with its own semantic model and RLS.

This is why shortcuts get described as one of the most consequential capabilities in Fabric — they deliver distribution, reuse, and a mesh-like topology without the operational tax of constant copying.

The Gotchas: Governance, RLS, and Access

Shortcuts don’t remove governance questions; they move them to the front door, which is where they belong.

Source permissions still apply — for internal sources the caller’s identity is used, so consumers may need read access to the source workspace, or you design sharing deliberately. Row-Level Security defined on a source table is enforced through the shortcut, which is correct behaviour but has a sharp edge: an RLS denial on one table can, in some configurations, break access to other tables in that shortcut for that user.

The patterns that hold up in practice: use shortcuts where security and RLS are consistent across the consuming audience; fall back to small curated physical copies where they aren’t; and document explicitly which datasets are shared via shortcuts versus internal-only. The takeaway for your team is short — shortcuts reduce data movement, not the need to decide who should see what.

External Stores, Security, and Direct Lake

The most strategically interesting direction is how shortcuts interact with external storage and Direct Lake. You can shortcut from OneLake into private ADLS Gen2 accounts using trusted workspace access, VNets, and private endpoints to keep traffic locked down. With the right configuration, you can run Direct Lake mode in Power BI over data that physically lives in external ADLS but is exposed through OneLake — low-latency reporting on a lake you already own. The same extends to S3, so OneLake behaves as a unified virtual lake across clouds rather than only inside Azure.

For an organization that already invested heavily in an ADLS- or S3-based platform, this is the pragmatic path: keep the existing lake, expose curated zones into Fabric via shortcuts, run Fabric workloads on top, and decide later whether any of it is worth re-platforming. Not rip and replace — plug in and evolve.

Hope for Lean Teams: Mesh Behavior Without Migration

A starting pattern that works without a dedicated platform team:

1. Pick one central gold lakehouse for your most trusted data — customers, products, financials.
2. Create domain workspaces (Sales, Operations, Finance) with their own lakehouses.
3. Shortcut from each domain into gold and, where needed, into external ADLS/S3 zones.
4. Build semantic models on the domain lakehouses, not on gold, so each domain evolves its own view without touching the canonical store.
5. Where security is tricky, use a small curated copy for that audience rather than forcing everything through shortcuts.

You get the mesh experience — ownership by domain, reuse of shared data, no mass migration — driven by design and configuration rather than endless data movement.

Where I Fit In

OneLake shortcuts won’t design your mesh for you. They give you the missing primitive: a way to share and reuse data across domains and clouds without turning your platform into a copy-and-paste factory. The architecture decisions around it — which datasets become canonical, where RLS flows cleanly versus where a curated copy is safer, how external lakes plug in — are where the real work sits.

If duplicated golds, orphaned pipelines, or a stalled migration plan are starting to weigh on your estate, and you want to redraw the architecture around shortcuts, let’s talk.

Isaac Truong | Founder, Allston Yale

Enterprise-grade analytics for $50M–$100M SMBs

Power BI | Fabric | Azure | Data Strategy

Book a 20-min Fabric diagnostic →

Subscribe to get Friday Fabric Facts in your inbox (plus early access to templates)

LinkedIn: Connect with me for daily Fabric tips

Friday Fabric Facts # 18: Originally Posted on LinkedIn, May 29, 2026