Forum Discussion
Online semantic model - import mode
Hi all,
I have a Lakehouse in Fabric, and a default semantic model. In this semantic model I have my relations between tables, my measures, etc...I love how I can manage this in one central place, as all dependent reports (in my case a few PBI reports, but also some connected Excel files) are all updated in the same time BUT...
my PBI reports connected to the semantic model are simply too slow SO...
I want to change to use import mode...
Is there a way that I can keep the benefits from the online, centrally managed semantic model, but still use import mode...let me clarify...
I want to use import mode in PBI. When I change or add a measure in the online semantic model, when I update a relationship between tables, when I create a new relationship between tables, and then refresh my PBI file, I want new measures to appear, changed measures to reflect the changes, new relationships to be reflected...so basically, this means a local copy of the online semantic model in my PBIX file, well aware that changes will not be reflected live, but will be reflected once I refresh my PBIX file...
From my experience, whenever I am connected to a semantic model, the connection is live. If I change it to import mode, I am only loading the tables, not the relations or the measures.
I would like to understand how to make such a connection if possible, or maybe there is another obvious solution I am not seeing here.
thx
Jan
Hi JS23 ,
First of all, my suggestion is not to use the default semantic model, there is a lot of problems with that specially because of the auto update that it gets when you add new tables, and all sort of things. I would create a new semantic model from the lakehouse and recreate the relationships and measures, you can do this using tabular editor or ALM toolkit to make it faster (XMLA needs to be turn on).
If you are using the model connecting to the Lakehouse this should be using the Direct Lake mode, except if you are hitting the guardrails of your capacity please refer to the image below
https://learn.microsoft.com/en-us/fabric/get-started/direct-lake-overview
The Direct Lake mode is also possible to edit using the Power BI desktop and the update is direct in the service and desktop, once again you need to have the XMLA turned on.
This model is very similar to import mode, so if you are having very slow reports something is happenning.
Altough if you still want to create an import model from your Lakehouse, and you can even get that imported model to be updated online however those updates online won't be automatically, done on the desktop version you have, for this you need to use an external tool like ALM Toolkit that allow you to do the merge between two different dataset's and that way you can have the changes matching.
1 Reply
- MFelix
Super User
Hi JS23 ,
First of all, my suggestion is not to use the default semantic model, there is a lot of problems with that specially because of the auto update that it gets when you add new tables, and all sort of things. I would create a new semantic model from the lakehouse and recreate the relationships and measures, you can do this using tabular editor or ALM toolkit to make it faster (XMLA needs to be turn on).
If you are using the model connecting to the Lakehouse this should be using the Direct Lake mode, except if you are hitting the guardrails of your capacity please refer to the image below
https://learn.microsoft.com/en-us/fabric/get-started/direct-lake-overview
The Direct Lake mode is also possible to edit using the Power BI desktop and the update is direct in the service and desktop, once again you need to have the XMLA turned on.
This model is very similar to import mode, so if you are having very slow reports something is happenning.
Altough if you still want to create an import model from your Lakehouse, and you can even get that imported model to be updated online however those updates online won't be automatically, done on the desktop version you have, for this you need to use an external tool like ALM Toolkit that allow you to do the merge between two different dataset's and that way you can have the changes matching.