AWS Outage Lasted 13 Hours⚠️|SAP pays $480M settlement💸|Time to Reassess MongoDB🤔

In partnership with

What’s in today’s newsletter:

• AWS AI tools cause 13-hour global cloud outage ⚠️

• SAP settles Teradata patent suit for $480M💸

• MongoDB's future hinges on innovation amid volatility 🤔

Also, check out the weekly Deep Dive - Microsoft Fabric OneLake

🤝 Sponsor Cloud Database Insider

Reach 3,400+ senior data architects, engineers, and platform leaders working with Snowflake, Databricks, PostgreSQL, and modern AI data stacks.

👉 Sponsor an upcoming issue

Want to get the most out of ChatGPT?

ChatGPT is a superpower if you know how to use it correctly.

Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.

Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.

Download the free guide

DEEP DIVE

A detailed look into Microsoft Fabric, Part Two, OneLake

Sometimes I wish Microsoft would keep things simple. But alas, I live in reality and my name isn't Satya, but my name is Gladstone.

Why do I say this. Well, When you hear the name OneLake, what comes to mind? Do you think file format? Do you think table format? Do you think some sort of storage stack. That is my point of saying it would be good if Microsoft kept their product naming simple.

I am still delving into the world of Microsoft Fabric and I even plan on writing the DP-700 exam in the latter portion of the year, but in the meantime to keep food on the table, I have to be knowledgeable about Fabric in general, and OneLake in particular.

So what is OneLake?

Let me put it as plainly as I can (because if Microsoft had their way, they’d probably call it “FabricUnifiedMultiEngineDeltaLakehouseStorageLayer v2.3 Enterprise Edition with Shortcuts™” and I’d still be confused).

OneLake is the single, logical, tenant-wide data lake that Microsoft Fabric creates for you automatically the second you turn Fabric on in your Microsoft 365 tenant. You don’t provision it. You don’t choose a region for it (workspaces are region-specific, OneLake itself is tenant-scoped). You can’t create a second one, and you sure as hell can’t delete the primary one. It just… exists. Like gravity. Or taxes. Or that one colleague who always replies-all.

Microsoft’s own marketing line — and for once I actually love it — is that OneLake is “the OneDrive for data.”

Just like OneDrive sits under your entire Microsoft 365 experience and lets Word, Excel, Teams, PowerPoint, and Copilot all work on the exact same files without you copying them around, OneLake sits under the entire Fabric platform and lets Data Factory, Spark (Data Engineering), Notebooks (Data Science), SQL warehouses, KQL databases (Real-Time Intelligence), Power BI semantic models (Direct Lake mode), and every future Fabric workload all read from and write to the exact same physical files with zero copying, zero ETL between engines, and zero “wait, which copy is the source of truth?” drama.

That last part is the entire revolution.

The Problem OneLake Was Born to Solve

Before Fabric you had:

• Raw data landing in ADLS Gen2 storage account A

• Spark jobs transforming it into storage account B

• Someone copying the good stuff into a dedicated Synapse dedicated SQL pool

• Power BI importing yet another copy into an import model

• Real-time team spinning up a separate Kusto cluster

• Finance team mirroring everything into Snowflake because “reasons”

Result? Data duplication explosion, massive egress bills, version angst, governance nightmares, and analysts arguing about whose numbers are right.

OneLake’s entire purpose is to say: “We’re done with that circus.”

💡 Enjoy deep dives like this?
Companies building developer and data infrastructure sponsor Cloud Database Insider to reach enterprise data teams.

👉 Partner with CDI

There is one copy of the data (stored in open Delta Lake format — Parquet files + transaction log). Every single Fabric experience talks directly to those files. Spark writes a Delta table? The SQL analytics endpoint in the lakehouse can query it instantly with T-SQL. Power BI can go Direct Lake on it with sub-second performance. KQL can external-table it for real-time. No pipelines moving data between experiences. No “refresh at 2 a.m. or the dashboard breaks” nonsense.

How It’s Actually Implemented

Under the hood, OneLake is Azure Data Lake Storage Gen2 on steroids with a massive abstraction layer on top. Microsoft calls the storage format “OneLake” but technically it’s still ADLS Gen2 with some very clever tenant-wide indexing and caching.

Every workspace you create lives inside OneLake. Inside workspaces you create items (lakehouses, warehouses, KQL databases, etc.). Those items are really just special folders in OneLake with extra metadata and engines attached.

All tabular data is forced into Delta Lake (open source, Apache-licensed). That gives you:

• ACID transactions

• Time travel (AS OF syntax in SQL, versionAsOf in Spark)

• Schema enforcement

• V-Order (Microsoft’s columnar optimization that makes Power BI Direct Lake incredibly fast)

• Liquid clustering instead of old-school partitioning

Unstructured and semi-structured files? Totally fine. Just drop them in. OneLake doesn’t care.

The Innovative Features of MS Fabric

1. Shortcuts
   Think symbolic links, but enterprise-grade and cross-cloud. You can shortcut:

   • Another lakehouse in a different workspace

   • An Azure SQL table

   • An entire S3 bucket

   • Google Cloud Storage

   • Dataverse tables

   • On-premises file shares via gateway

   • Even Iceberg tables from Snowflake (with write-back in preview)

   The data never moves. It just appears inside your lakehouse as if it lives there. Your Spark job, SQL query, and Power BI report all see it natively. Caching is configurable (1–28 days) so you’re not hammering the source every time.

2. Mirroring
   Near-real-time continuous replication from Azure SQL DB, Snowflake, and (coming) more sources. Microsoft gives you the mirrored storage for “free” up to your capacity limit. Mind blown.

3. OneLake Catalog
   The governance brain. Search across every workspace in the tenant, see lineage, apply sensitivity labels, set folder-level RBAC (preview but rolling out hard in 2026), and get Copilot-style recommendations.

4. OneLake file explorer + Windows sync client
   Yes, you can literally map OneLake as a network drive on your laptop. I did it last week and felt like a wizard.

Who Is This Actually For?

If you’re a 50-person startup running everything in one Snowflake account… maybe not yet.
But if you’re a mid-to-large enterprise with:

• Multiple business units

• Multi-cloud or hybrid data estate

• Heavy Power BI usage

• Data engineers, analysts, scientists, and citizen developers all fighting over the same data

…then OneLake is the thing that finally makes the “single source of truth” dream stop being a lie your CDO tells in town halls.

I’m writing this while studying for DP-700, and the more I play with it in my personal tenant, the more I realize: Microsoft didn’t just build another data lake. They built the operating system for enterprise data. OneLake is the file system. Fabric experiences are the applications. The capacity units are the CPU/RAM. And shortcuts + mirroring are the universal USB-C ports that let you plug in the rest of the world without conversion cables.

Next week I’ll go deeper into the medallion architecture inside OneLake (bronze/silver/gold done right).

But for now, just know this: when someone asks you “What is OneLake?” you can smile and say:

“It’s the reason you’ll never have to explain to your CFO why you have 14 copies of the same 2 TB dataset ever again.”

And if that doesn’t make you a little bit excited… well, maybe your name actually is Satya.

Gladstone Benjamin

🚀 Work With Cloud Database Insider

Looking to reach enterprise data engineers and architects?

Limited sponsorship slots available each month.

👉 Sponsor Cloud Database Insider