Clearing the Confusion Around Microsoft Fabric

In this blog, we cover:

1. What is Microsoft Fabric?↵
2. What is OneLake? ↵
3. What if I’m already using Power BI, ADF, or Synapse (which are now components of Fabric)? ↵
4. What does Fabric add to Power BI?↵
5. What is the licensing structure for Fabric? ↵
6. What other advice do you have for those considering Fabric?↵

What is Microsoft Fabric?

Fabric is a cloud SaaS (software as a service) product that brings together several services — most of which have existed for some time — into a single, end-to-end data analytics platform.

From Microsoft’s perspective, Fabric is an incentive for its users to adopt more of the Microsoft Azure data ecosystem, because it brings Azure Data Factory, Azure Synapse, and Power BI together into a single product. It integrates workflows and encourages data democratization from data engineers to data analysts to business users.

The product allows users with seven different components, or “experiences”:

1. Data Engineering: Notebook-driven tool for transforming data at scale using Spark.
2. Data Factory: Tool for data engineers to build integrations from a more than 200 native connectors. Combines the simplicity of PowerQuery with the scale of Azure Data Factory.
3. Data Science: Experience to build, deploy, and operationalize ML models integrated with Azure Machine Learning.
4. Data Warehouse: Delta Lake columnar storage engine for data providing a SQL endpoint, fully separating compute from storage.
5. Real Time Analytics: Experience for streaming data (often unstructured or semi-structured) from an Azure event hub and storage in KQL databases.
6. Power BI: Experience for Data analytics and data visualization.
7. (NEW) Data Activator: No code platform to detect actionable conditions and trigger actions (think Teams messages, emails, PowerAutomate tasks).

What is OneLake?

The foundation on which all Fabric services are built is a data lake called OneLake (also new to the scene). Think OneDrive for data. OneLake provides a SaaS experience built on top of Azure Data Lake Storage (ADLS Gen2). It provides a unified location to store all organizational data where all the above experiences operate.

OneLake is also designed to allow mounting of existing PaaS storage accounts using a Shortcut feature – without the need to migrate existing data. Shortcuts can be connected to existing data sources in ADLS Gen2 or AWS S3 buckets.

OneLake is intended to eliminate ‘data silos’ for the democratization of data. While the data assets are centralized, each developer or business unit can create their own workspace, ingest the data into their lakehouses, process, and analyze data. This architectural pattern provides one source of centralized data, which enables decentralized or distributed teams.

What if I’m already using Power BI, ADF, or Synapse (which are now components of Fabric)?

The good news is none of those platforms in isolation are going away. They will be available for years to come.

If you have existing Power BI workspaces

You can turn on a Fabric capacity, no problem. There is no migration involved there. Your content still lives in the workspace, and you have those additional experiences at your disposal. Power BI still offers the same capabilities as before — you can build datasets using import or DirectQuery, build reports, publish reports, etc. Power BI Desktop is also not going anywhere. The Power BI experience won’t fundamentally change.

If you have Power BI projects in flight or planned for this year

This is not an obligation to change course to fit the new Fabric paradigms. But now that you have these new features available in your toolkit, you may want to consider them as options for the future.

If you’re using ADF or Synapse

There is no direct or automatic path to migrate today. ADF pipelines cannot be directly migrated to Fabric — though this is on the roadmap. Synapse spark workloads are easily migrated, but the underlying data still must be connected to OneLake.

The key differentiator for Synapse in Fabric is Azure Synapse is a PaaS whereas Fabric Synapse capability is SaaS, so moving from Synapse to Fabric will be fundamentally different. Fabric is considered a successor to ADF and Synapse and is intended to enhance the experiences from those platforms. This takes away the need to manage several Azure resources.

Provisioning Spark clusters will provide a significantly faster and easier experience. The compute resources you purchase are pooled with your Power BI resources. This may be perceived as a benefit to some and a detriment to others who feel it will increase total cost of ownership (TCO).

However, managing technical debt is challenging to quantify. Fabric can simplify and centralize your codebases — and that is difficult to put a price tag on.

What does Fabric add to Power BI?

Fabric will enhance the overall developer experience, specifically with Git integration features.

The inability to use Git for .pbix files because they are stored as compressed binary files has long been a pain point for Power BI developers.

Fabric is bringing more source control ability to workspaces. Workspaces can be linked with Azure DevOps, where the underlying data models and reports can be accessed as JSON files in a repository. With this, Power BI development is moving toward software development paradigms — models and reports can be compared in diffs, merged, and changes approved in Pull Requests.

(As of July 2023, only reports and datasets support this Git integration.)

What is the licensing structure for Fabric? 

There are two licensing components of Power BI that are critical to understanding the overall licensing model: user-based licensing and capacity-based licensing.

User-based licenses are not changing with the introduction of Fabric. Everyone in your tenant can be assigned a free license, a Pro license, or a Premium Per User License (PPU). Without going into depth of the individual features, this is basically what they mean:

• Free licenses provide an entry point to publish reports to a personal workspace on the web with limited sharing capabilities.
• A Pro license is an entry point to creating workspaces for collaboration with other Pro users.
• A PPU license is an elevated Pro license as an entry point to Power BI Premium features for dataflows, Datamarts, using the XMLA endpoint, and more.

Capacity-based licenses provide dedicated pools of resources measured in Compute Units (CU) (CPU, memory, etc.) for hosting your workspaces. There are three different types of SKUs for capacities:

1. Embedded SKUs are generally for independent software vendors who want to embed Power BI content into web applications. Embedded SKUs are deployed via the Azure Portal — opposed to through the M365 portal for others. If you are on an embedded SKU, it does not support Microsoft Fabric — so don’t anticipate any changes there.
2. Premium SKUs are generally used for large enterprises to support large dataset volumes. The smallest Premium SKU starts at around $5,000/month and requires an annual commitment. One of the major benefits of a Premium capacity is additional sharing capability. Premium workspace applications can be shared with users on a free license. Those on P SKUs will be given access to Fabric features and will become Fabric SKUs themselves in the future.
3. Fabric SKUs have a pay-as-you-go pricing model and is billed by the second. Fabric SKUs can be paused as well to limit costs. They are much more accessible for experimentation — the lowest Premium SKU (P1) provides 32x the resource pool of the lowest F SKU (F2). An F2 provides an entry point for experimenting with Fabric and would cost less than $300/month without pause. Fabric SKUs provide the ability to scale up or scale down your capacity as necessary. When you scale up an F SKU to F64, you enable the same sharing feature as a Premium Capacity.

Do you need to choose between a user-based and a capacity-based license?

Not necessarily. User-based licenses can be used without a capacity-based license, but capacity-based licenses cannot be used without user-based licenses.

Any developer who wants to use the workspaces in a dedicated capacity must have at least a pro license. With that in mind, not all F SKUs are created equal — they do not all provide the same sharing capability. F SKUs from F2-F32 requires Pro licenses from anyone who want to develop or engage with the content in a workspace. F64 and above is more appropriate for larger organizations when buying hundreds (or thousands) of individual licenses is not feasible.

Note: An additional capacity-based license — reserved instance (RI SKU) — which will be coming soon, will provide the same experiences as Fabric (F SKUs). The key difference is in its billing structure – billed monthly based on an annual commitment like a P SKU. An annual commitment provides some cost saving benefits. But since we don’t know exactly when that will come or how much it will cost, it’s difficult to plan for.

What other advice do you have for those considering Fabric?

1. Remember that — as with any platform or technology — Fabric is not a silver bullet.

Opening your wallet for Fabric won’t immediately revolutionize your data analytics capabilities. There’s a people and a process component that can’t be ignored.

You need:

• Proper training and enablement for your developers.
• A Center of Excellence (CoE) to ensure that new features are used properly.
• Data governance, especially if you work with sensitive data.
• A data strategy — there are limitless ways you could architect a solution in Fabric — consider what works best for your organization.

2. It’s new, so there are some gaps (as of July 2023):

• Because Fabric is largely a GUI-driven SaaS platform you may experience some frustration with the navigation prior to its GA release.
• Best practices are lacking with the introduction of OneLake and workspaces in a medallion architecture.
• The governance model in the new SaaS platform will likely involve some growing pains.
• There isn’t currently a way to isolate workloads in a Fabric cluster.

If you’re not sure what value Fabric will bring to your use cases, try it out — there is no commitment. With a free PBI license you can start a 60-day free trial of Fabric. Consider how its features would fit into your data architecture needs. You can still elect to use Power BI exclusively.

For organizations new to Power BI — perhaps those with some Pro licenses — Fabric is still very accessible from a cost perspective even beyond the free trial. You can get access to everything Power BI already is along with ADF and Synapse Data engineering capability.