The following schema summarizes the architecture of Power BI 2.0:
In this picture, we can see from left to right, the sources, service and destinations:
- The sources: there are three types of sources in Power BI: files, databases and services
- The service: you can define three types of objects: datasets, reports and dashboards
- The destinations: dashboards and reports can be consumed in three ways,
You can see the data sources in this picture:
They are separated in three categories:
- Files (which formats are Excel and Power BI)
- Services, which are prepackaged models of data for popular services (like Google Analytics or GitHub)
- Big Data and More (for Azure DBs and SQL Server Analysis on-premises)
Power BI 2.0 supports a new direct connectivity to Apache Spark (especially suited for Big Data scenarios) which Microsoft previously announced. Query performance over a Hadoop dataset can be 100 times faster with Spark.
The three main concepts of Power BI are:
A dataset is basically a set of tables. Each table can be the result of a Power Query. All tables of in a dataset can be completed with relationships, measures, etc.
You can combine several datasets in a report and again you can combine visualizations from several reports on a dashboard (this can’t be done in Power BI Designer but only in the service).
One of the best features of Power BI is that everything you can do in the service is accessible via the API, which is described here: Overview of Power BI REST API.
You have access to three types of objects:
You can use PowerShell to inject data in real time to your datasets.
For example, this GitHub project has defined a very quick and dirty sets of cmdlets and especially the Out-PowerBI cmdlet:
You can find Patrick's other articles in this series on [ #Office365] Power BI 2.0:
- The big picture
- Architectural aspects (this article)
- More on data sources
- More on Power BI mobile apps
- The gateways