This week at Microsoft Build, we announced the future.

With an introduction like that, I should probably remind everyone that this is my personal blog, my personal perspective, and my personal opinions. Although I am a Microsoft employee, I am not speaking for or otherwise representing my employer with this post or anything else on this blog.

With that disclaimer out of the way, let’s get back to the future. Let’s get back to Microsoft Fabric.

According to the official documentation, “Microsoft Fabric is an all-in-one analytics solution for enterprises that covers everything from data movement to data science, real-time analytics, and business intelligence.” Fabric is implemented as a SaaS service for all sorts of data, analytics, and BI capabilities, including:

• Data integration
• Business intelligence
• Synapse data engineering
• Data warehousing
• Data science
• Real-time analytics

I’ve been working on Fabric for around 18 months^[1], and I could not be more excited to finally share it with the world. I don’t own any of the features coming in Fabric, but my team and I have been running an NDA private preview program with thousands of users from hundreds of customer organizations around the world building solutions using Fabric, and providing feedback to the product team.

This introductory blog post won’t attempt to be super technical or comprehensive. Instead, I’m going to share the information I’ve shared most frequently and consistently during the Fabric private preview – the information and context that will help you get started, and help put that more technical information into context.

For folks who are already familiar with Power BI^[2], Fabric is going to feel familiar from day one. This is because the SaaS foundation on which Fabric is built is the Power BI service you already use every day.

The foundation is evolving and improving, and there are new capabilities in lots of places, but the foundation of Fabric is the foundation of Power BI. This means that from day one you already know how to use it:

• Workspaces – Fabric workspaces behave like Power BI workspaces, but with more item types available.
• Navigation – If you know how to move around the Power BI portal you know how to move around the Fabric portal, because it works the same way.
• Collaboration and content management – You can collaborate and share with Fabric items and workspaces just like you do with Power BI.
• Capacities – New Fabric workloads use the capacity-based compute model used by Power BI Premium. If you don’t already have a capacity, you can start a free trial.
• Administration – Fabric administration works like Power BI administration, and the Fabric admin portal is the evolution of the Power BI admin portal. To enable the Fabric preview in your Power BI tenant or for a specific capacity, you can use the admin portal.
• Much, much more – I won’t try to list everything here, because there’s already so much documentation available.

At this point you probably get the idea. If you’re familiar with Power BI, you’re going to have an easy time getting used to Fabric. Power BI will continue to evolve and grow, and there are a lot of exciting improvements coming to Power BI in Fabric^[3] even without taking the new capabilities into account.

But what about those new capabilities? What about all the new data integration, data science, data engineering, data warehousing, and real-time analytics capabilities? How familiar will they be?

That’s a slightly more complicated question. In a lot of ways these new Fabric workloads are the evolution of existing Azure data services, including Azure Synapse, Azure Data Factory, and Azure Data Explorer. These established PaaS services have been updated and enhanced to run on the Fabric shared SaaS foundation, and their user experiences have been integrated into the Fabric portal.

If you’re already familiar with Azure Synapse, Azure Data Factory, and/or Azure Data Explorer, the new capabilities in Fabric will probably be familiar too. You already know how to work with pipelines and notebooks, and you already know how to write SQL and KQL queries – in Fabric you’ll just be doing these familiar things in a new context.

There are a few key Fabric concepts that I’ve seen more new-to-Power BI preview customers as questions about. If you or your colleagues are more Azure-savvy than Power-savvy, you’ll probably want to pay attention to:

• Capacities – Fabric uses capacities for compute across all experiences^[4], which provides a consistent billing and consumption model, but which will necessitate a change in thinking for folks who are used to other service-specific approaches to billing and consumption.
• Workspaces – Other services don’t have the same concept of a workspace as Power BI and Fabric do… but some of them have different concepts with the same name. Since workspaces are a crucial tool for content creation, organization, and security, understanding them and how they work will be important for success with Fabric.
• A “managed” SaaS data service – In most data services, the “catalog” of items and their relationships is expressed through the metadata of a given instance. This means that capabilities like discovery, lineage, impact analysis are either absent, limited in scope, or only available through integration with an external data catalog or similar service. Fabric, like Power BI, maintains an internal data catalog for all items in the tenant, and their relationships to each other. This information is exposed through APIs and integrated into experiences like the workspace lineage view and the data hub, making it easier to discover, understand, and use data.

In addition to things in Fabric that will be familiar to people with Power BI experience and things in Fabric that will be familiar to people with Azure data experience, there’s one huge part of Fabric that is going to be new to everyone: OneLake.

OneLake is a SaaS data lake that is a key component of the Fabric SaaS foundation^[5]. Every Fabric tenant includes a single OneLake instance, and all Fabric experiences work natively with data in the lake.

• OneLake is open – OneLake is built on ADLS Gen2. You can store any type of file, and use the same APIs you use when connecting to ADLS Gen2. Storing data in OneLake doesn’t mean it’s locked into Fabric – it means it can be used where and how you need it to be used.
• Delta by default – Fabric experiences store their data in OneLake in parquet delta files. Delta is an open source, compressed columnar format that supports ACID transactions, and is supported by a wide range of tools.
• Store once, use everywhere – Because there’s one OneLake, data can be ingested and stored once and used where it’s needed. You can have a single set of delta files that are exposed as a lakehouse and manipulated using notebooks, while at the same time are exposed as a warehouse and manipulated using SQL, and exposed as a Power BI tabular dataset in DirectLake mode. This decoupling of storage and compute is enabled by OneLake, and I expect it to be one of the most significantly game-changing aspects of Fabric as a whole.
• OneLake is integrated – Being open makes it easy for you to store your data in OneLake while using it with whatever tools and compute engines you choose. OneLake shortcuts allow you to keep your data where you have it today while logically exposing it as if it were stored in OneLake.

OneLake takes the familiar concept of a data lake, and puts it where no one sems to expect it: at the center of work, where it makes sense, deeply integrated into the tools and experiences used by everyone contributing to a project or product.

With all of these new and familiar capabilities coming together into a single SaaS platform, the next thing that Fabric delivers is a comprehensive set of user experiences.

Modern data projects often involve a wide range of practitioners – data scientists, data engineers, data developers, BI developers, report authors, and more. Before Fabric, each practitioner persona would typically work in their own set of tools and experiences, each of which had its own strengths and weaknesses and capabilities. When taken together, this means that most projects involve significant integration effort to make the output of one tool work with the next tools in the value chain – and often there are tradeoffs made to accommodate the mismatch between tools.

With Fabric, each task and persona has a purpose-built set of experiences that all work natively with the same data in OneLake. The result is that data practitioners can focus on delivering value through their data work.. not on building integrations so their tools will work together. Teams can set up workspaces that contain the  data and items they need – lakehouses, warehouses, notebooks, spark jobs, pipelines, dataflows, dataset, reports, and more. Data in one workspace can be used in other workspaces as needed, and because of OneLake it can be stored once and used multiple times without duplication.

During the Fabric private preview, the chief data officer of a well-known global organization^[6] said something to the effect of:

With Fabric I can finally be a Chief Data Officer instead of being a Chief Integration Officer.

And this is why I believe Fabric represents the future of data.

Think back 10-12 years when the first generation of PaaS data services were becoming available. Many data practitioners looked at them and dismissed them as solutions to imaginary problems – why would we ever need a cloud service when we had these beautiful database servers in our own data centers, with IO subsystems we’ve designed to our own specs and fine-tuned to the nth degree^[7]? It took time for people to realize the value and advantage of the cloud, but today there are entire classes of problems that simply don’t exist anymore because of the cloud.

I believe that the integrated, open, flexible, SaaS nature of Fabric means that we’re at an inflection point as significant for data as the advent of the cloud. Fabric will eliminate entire classes of problems that we take for granted today – in a few years we will take this new platform and this new paradigm for granted, and wonder how we ever thought those problems were an acceptable part of our professional lives.

Welcome to Fabric. Welcome to the future of data.

Ok, that’s my post. Where should you go from here? In addition to all of the links above, you should definitely check out the official blog post for the official big picture. You should also join us Wednesday and Thursday for a “simulive” virtual event as we go deeper into many of the key capabilities now available in Fabric.

I’ll see you there.

---

^[1] Old man voice: Back when I was your age we called Fabric “Trident” and we weren’t allowed to talk about it in public because if the Kaiser heard about it our boys fighting in France would be at risk! Let me tell you about the time I…

^[2] If you’re reading this post on this blog, I suspect this includes you. I’d love to know if you agree with my “feels familiar from day one” assertion.

^[3] I’m working under the assumption that the interwebs will be flooded today with blogs and announcements and guys in cubes, so I’ll leave it up to you to track down what’s exciting for you.

^[4] Yes, you need a capacity for all new Fabric experiences. Power BI licensing is not changing, but to work with the new Fabric capabilities you need a capacity to run them on. Fabric capacities are available in smaller SKUs than Power BI capacities.

^[5] You probably noticed it in the diagram image above, sitting there in the middle all integrated and important.

^[6] You know this company, but since this is my personal blog I’m probably not going to get their permission to name them. Also, as I write this blog post I can’t find the verbatim customer quote, so you’ll need to rely on my imperfect memory for this one.

^[7] As I type this in 2023, I can’t remember the last time I worked with an on-prem production database. It was probably 2011 or 2012. It feels like something from another age, another life.