Forum Discussion
Measures and CalcColumns in Power BI vs Tabular model (replace SSAS with Power BI?)
Hello,
I have a data-mart on-premise, and on top of that I have an ssas tabular model.
I've been playing with Power BI on my deskop and I noticed how it allow us to create our own calculated columns and measures. If that is the case then I am wondering if a tabular model is at all necessary. I mean, couldn't I simply connect straight into the data-mart and add all measures and calculated columns within Power BI instead?
I am sure that this might be the case of undocumented features and limitations. So I was hoping if someone else could help me shed some light on this?
Thanks for your help.
Regards,
P.
You need Analysis Services if you have one of the following requirements:
- You need to have row level security
- Your data set is greater than 250MB
- You need to have a fine grained control over how your model is refreshed from the source data (partitioning)
If the aforementioned requirements are not present, then using Power Pivot or Power BI Desktop to build your model and publish it to Power BI is a valid and wonderful choice.
9 Replies
pmdci,
The value of the tabular model is that you can directly connect to it from the Power BI service. Create your reports up there, and its a live connection that automatically refreshes whenever you update your model on the back end.
The desktop designer can be used as your replacement for the tabular model, but if you already have one, you are just adding an extra step. I build my report visualizations on the site.
The AS Connector can be downloaded from your app.powerbi.com site (click the down arrow in the upper right)
More info:
Hi Eno1978,
I think you misunderstood my question.
What I am trying to find out is if I can simply replace the SSAS tabular model by creating measures, calculated columns and relationships in Power BI instead -- thus making SSAS irrelevant.
Bottom line: If I could replace SSAS with a cloud-only feature, I would.
pmdci the Power BI desktop is not a "cloud only" feature, and the site doesn't allow you to build models.
Regardless, Could you move your model into the desktop tool? "Sure". Here are the trade-offs.
SSAS - row level security (Desktop - none)
SSAS - Does not store you information in the cloud (Desktop will)
SSAS - Live connection to your model (Desktop - only update up to 8 times a day)
I would not compare the modeling capability in Power BI directly to SSAS as equal in performance and capabilities. In terms of the tools, I would say (in general) the modeling tool in the desktop tool is designed for small projects, and SSAS is best utilized in larger company wide implementations. (especially if you have the option).
I don't see why not, you will need the Power BI SSAS connector, which supports row level security, here is more info:
Once you suck information into Power BI, or Excel PowerPivot for that matter, it is in a tabular model, so not sure why you would need another tabular model in between.
Hi smoupre,
I think you misunderstood my question.
My question is whether I could simply bypass SSAS, connect straight into the relation DBs and have all measures, calculated columns and relationships that exist in the SSAS created in Power BI instead. Therefore rendering SSAS irrelevant.
Regards,
P.
You need Analysis Services if you have one of the following requirements:
- You need to have row level security
- Your data set is greater than 250MB
- You need to have a fine grained control over how your model is refreshed from the source data (partitioning)
If the aforementioned requirements are not present, then using Power Pivot or Power BI Desktop to build your model and publish it to Power BI is a valid and wonderful choice.
Thanks everybody for their input.
Just to clarify: The rationale behind my question is whether I can remove SSAS server and have all the calculated columns, measures relationships and so on created using the Power BI desktop designer instead. This way Power BI would connect straight into the relational DB (the data mart).
My understanding from Andre's answer is that the modelling features of Power BI have some limitations:
- No row level security
- Model cannot be greater than 250MB
- No fine grained control over how model is refreshed from the source data (partitioning)
Am I right to assume that if this limitations are overcome by MSFT, then an SSAS server would no longer be necessary in my case?
Regards,
P.
PS: Apologies for the late response. I got no email alerting me of replies into the thread.
Thanks Andre!
Actually I got all confused due to the overlap of BI products by Microsoft. But one thing that put it all into perspective for me is that Power BI's modelling features are like PowerPivot in Excel. Once I got that in my mind, then it all made sense.
It is simply "PowerPivot for Power BI" -- for the lack of a better term.
Regards,
P.
