Forum Discussion
Helper Vingesting data from on premise sql server to Fabric Via Notebook
Power ParticipantHi,
I don't think the dataflows are the ones to be blamed here, although this is a very specific experience and it would be interesting to hear from someone from Microsoft.
Dataflows are linked to the Fabric capacity as the notebooks, the performance should not have this huge difference, it should be only a matter of choice - I believe so.
Anyway, the bottleneck on this case doesn't seems to be the dataflow execution, but the transfer through data gateway, which notebooks don't support yet.
First, I would suggest to analyse the volume of data: would it be possible to make incremental load to reduce it?
Second, I would consider alternate ways. What if you make some automation on premises, may be SSIS, to generate files and upload these files to the "Files" area of a lakehouse, than start in the cloud from this point?
At the moment you may be tied by the lack of notebook support to the data gateway. In the future, you may need a support ticket to ask about the best option and performance issues.
Kind Regards,
Dennes
Thanks for your explaination here Dennes. This issue bother me as well for these days.
Now I have to thinking use dataflow to ingest source data instead of Notebook since it does not support data gateway.
Would like to add one point : you know sometimes the source data tables could be hundreds or thousands and I need to manually config the destination one by one in dataflow as below.
In addtion, notebook will be more easier to mainten if anything need to be changed, like datasource connection string change., data base tables change, etc.