Prologika Newsletter Fall 2018
Semantics relates to discovering the meaning of the message behind the words. In the context of data and BI, semantics represents the user’s perspective of data: how the end user views the data to derive knowledge from it. A modeler translates the machine-friendly database structures and terminology into a user-friendly semantic model that describes the business problems to be solved. To address this need, you create a semantic model. In my “Why Semantic Layer?” newsletter I explained the advantages of an organizational semantic model. In this newsletter, I’ll discuss how Power BI extends semantic modeling and converges it on a single platform. But before I go into details and speaking of semantic models, I’m excited to announce the availability of my new “Applied DAX with Power BI” workshop and its first in-person and public enrollment class on Oct 15-16 in Atlanta! Space is limited so RSVP today.
Semantic Model Flavors
In Microsoft BI, you can implement a semantic model using Power BI Desktop, Excel (Power Pivot) and Analysis Services (Multidimensional and Tabular). The first two are typically used by data analysts, while Analysis Services is considered a professional technology. Thus, we can classify semantic models into two broad categories: personal (self-service) and organizational. Behind the scenes, Power BI Desktop, Power Pivot and Analysis Services Tabular use the same foundation and storage engine. Nevertheless, up to now there have been feature differences and a strict division between these two types.
| Personal | Organizational | |
| Author | Data analyst | BI Pro |
| Tool | Power BI Desktop, Excel (Power Pivot) | SSDT and Analysis Services |
| Scope | Narrow (usually to solve specific need) | Wide (multiple subject areas) |
| Implementation effort | Short | Longer (data warehouse, ETL, model, testing) |
| Data capacity | Limited (up to a few million rows) | Larger (millions or billions of rows) |
| Data quality | Trust author | Trust modeler and testers |
| Data centralization | May lead to data duplication | Single version of truth |
| Deployment | Power BI Service, Power BI Report Server | SSAS (on premises) Azure Analysis Services (cloud) |
| Consumers | Department, project | Potentially entire organization |
How Power BI Changes Semantic Modeling
Power BI will blur the boundary between the personal and organizational aspects of semantic modeling. First, we’ve already seen how Microsoft introduced the following “pro” features in Power BI that don’t even exist or more difficult to implement with Analysis Services:
- Incremental refresh – Delivers the ability to refresh portions of a larger dataset, such as the last 7 days. Currently, the largest dataset size supported by Power BI Premium is 10 GB, but Microsoft has already announced that soon Power BI will support larger datasets. What this means for you is that you’d be able to deploy organizational semantic models to Power BI Premium and schedule them for incremental refresh. My blog “Notes on Power BI Incremental Refresh” provides the details on this feature.
- Composite semantic models – A composite model has heterogenous storage, such as some tables are imported and some are DirectQuery, as I discussed in my blog “Power BI Composite Models: The Good, The Bad, The Ugly“. This brings a lot of flexibility to how you connect to data.
- Aggregations – Aggregations are predefined data summaries to speed up queries with very large models. My blog “A First Look at Power BI Aggregations” covers Power BI aggregations.
On the tooling side of things, Power BI Desktop will also pick “pro” features, such as perspectives and displays folders. Microsoft hopes that in time Power BI Desktop will serve the needs of both data analysts and BI pros. However, the lack of extensibility and source control, as well as performance issues caused by committing every model change to the background Analysis Services instance, makes me skeptical that pros will embrace Power BI Desktop. But because Microsoft announced plans to open the Power BI Tabular management endpoint (recall that published Power BI datasets are hosted in a “hidden” Tabular server), pros can still use SSDT and community tools, such as Tabular Editor, to design and deploy their models to Power BI Premium.
In time Power BI Premium will become a single cloud platform for hosting Microsoft BI artifacts (semantic models and reports) and facilitating the continuum from personal to organizational BI. This is a great news for BI practitioners frustrated by tooling and deployment differences. At the end, the personal and organizational paths will converge without feature discrepancies. The only difference would be the scope of your organizational model and how for you want it to become “organizational”.
Teo Lachev
Prologika, LLC | Making Sense of Data
Microsoft Partner | Gold Data Analytics
