Metadata in Power BI

Question

Years ago I used a BI product called Hyperion Interactive Reporting. It allowed me to connect to a data source and create data models from which I would create reports. So far, sounds like Power BI right?

It also had the ability to connect to a metadata repository database. This database would contain data that mapped the actual, often cryptic, table and column names in the database to human-readable, business terms. For example a column that I saw in Hyperion as "Cost Center" may have been in the database as costCenter, work_order, or PROJECT-NUMBER. (It would also allow me to define the default join paths, but let's keep this question smaller.) This provided a way to make report development easier.

In Power BI, I see that I can manually rename columns, one-at-a-time. (And each time I touch something minor like this, Power BI takes several seconds to validate the entire file.) But I also see the need for many models that use the same data sources. So I may be defining the "Cost Center" column a few hundred times (a handful of reports per data set to answer a specific type of question, a few data sets that need Cost Center because the transformations in the model will be different for each type of question, several different combinations of data sources that include Cost Center, etc.)

Is there a way to connect Power BI to a metadata repository? Is there a way in Power BI to say, "Across all of my models/datasets, if I'm using the costCenter column from the Financial database, display Cost Center to the user"?

With about 20,000 columns in my data warehouse and 20,000 reports in my current reporting system, this could become a big deal if we intend to migrate to Power BI.

Answer 1

TLDR; There isn't an easy way to achieve this. What you have now is probably better than you could achieve without a ton of work. If you do try it, use SSAS instead of Power BI Desktop to author models.

Does Power BI have a metadata repository? No. There are tools that can get metadata from Power BI models, but you would have to manually build the metadata repository. If you want a centrally managed environment like this, I would highly recommend using SQL Server Analysis Services (SSAS) for on premise, or even better, Azure SSAS in the cloud. (Azure SSAS will get new features sooner than SSAS installed on premise.) While Power BI Desktop is a great self-service tool, I wouldn't author in it if I needed to control or report across the environment. There just aren't easy ways to corral all of the Power BI models in a report and it's a much more manual process. SSAS will need IT Support and is a higher cost and you will hit more issues than Power BI Desktop, but you will need it if you need central control. It's possible that more management controls will be added to the PowerBI.com service over time, but as of November 2021, you can't do this easily.

So what's the difference between Power BI Desktop and SSAS? The same Power BI engine in Power BI Desktop also exists in SSAS. When you start Power BI Desktop, it's actually starting a SSAS instance behind the scenes. Using SSAS directly just makes it easier for you to connect to the database behind the scenes and see all the models in the environment from one place, while Power BI Desktop doesn't let you peak behind the scenes and it only loads a single model at a time.

How do you get the metadata? It is an easy thing to get SSAS metadata using Power Query (or any SQL tool) to pull Direct Management View (DMV) data. DMVs are management tables that hold all of the metadata of the model, and you just use SQL commands to get the data. Search on "SSAS DMV" to learn more. I have a Excel file that uses Power Query to pull all the key DMV views for all our models in our servers. It makes it easy to do the kind of analysis as in your example.

For Power BI Desktop, you can connect to the hidden SSAS instance and do the same thing, but the report has to be open to do it, and there is no easy way to refresh the data--you pretty much just repeat the process each time. You will connect via localhost:port_number, and the port number is randomly created each time you start Power BI making it impossible to refresh the data pull. There are External Tools such as DAX Studio, Power BI Helper, and dataMarc's Document Model that make that easier, but there's no easy way to automate building the metadata repository for Power BI Desktop files. I would use SSAS directly rather than trying to automate building a large metadata repository.

What about making changes to models? To my knowledge, there aren't any tools that make it easy to make changes across models, though again, you could manually build them. I don't think I would trust my own tool to automate changes across models though. There's just too much that could go wrong. But if you had the need and the budget, you could build it. Look at tools like Tabular Editor, ALM Toolkit, and Microsoft's SSMS, and read on DevOps pipelines for automating updates. These tools work against SSAS and Power BI Desktop, but again, you have to open the Power BI files to work with those models, which makes automation that much harder to do.

Note that all the external tools I've mentioned except Tabular Editor v3 are free (though Tabular Editor v2 is free). PowerBI.tips is a great place to install all these tools from a single installer.