A significant barrier to leveraging a large Power BI dataset is the link between Power BI Live connection reports and the dataset(s) in their app workspace. For example, if an organization wishes to create three Power BI apps targeting three separate business units, a separate Power BI dataset currently needs to be hosted within the app workspace for each app. Naturally, any BI/IT organization would want to avoid creating (and managing) copies of a Power BI dataset, including its data model, queries, and measures.
This barrier or limitation is being worked on by the Power BI team so that, at some point in the future, a single Power BI dataset is expected to be able to support reports hosted in multiple workspaces. This isolation between visualizations (reports and dashboards) and datasets would lend itself to the following solution architecture:
As shown in the preceding diagram, a team's dataset designer(s) could be responsible for a single but potentially very large and complex Power BI dataset, including many fact and dimension tables. The dataset would be hosted in a premium capacity with v-cores and RAM aligned to its size and incremental data refresh could be configured to only load recent or new data. From a data visualization and distribution standpoint, report authors would create Live connection Power BI reports based on the large dataset (AdWorks Enterprise) and publish these reports to specific app workspaces. These reports would be refreshed based on the refresh schedule for the source dataset and, in the case of DirectQuery datasets, a scheduled cache refresh could be configured for the dashboards.
As of February 2018, no timeline has been provided on the availability of this feature. Therefore, even with support for incremental data refresh and very large datasets, the need to maintain a single, consolidated data model will lead many organizations to choose Analysis Services models for corporate BI solutions. Additional reasons for choosing Analysis Services models over Power BI Desktop are included in the Analysis Services section later in this chapter.