Forum Discussion
Optimize Service memory usage
DouweMeer So when you are using Import mode for your tables, whenever the report is used, the dataset is loaded into the capacity's memory. When the refresh happens, roughly the dataset size x2 is reserved in the memory.
On the other hand when using DirectQuery, only the materialized results are cached in the memory.
We did some tests by uploading a second report non 'used' version of the same. During report usages and refresh we didn't saw a behavior of what you're suggesting. Is there any documentation on the basis what you're referring to?
Note for the Direct Query would be that you would put stress on the server capacity by making requests each time it is used. We don't know how both would size up when compared to each other.
- JirkaZ6 years ago
Solution Specialist
DouweMeer I would start with these two docs.
Import: https://docs.microsoft.com/en-us/power-bi/guidance/import-modeling-data-reduction
DirectQuery: https://docs.microsoft.com/en-us/power-bi/guidance/directquery-model-guidance
- DouweMeer6 years ago
Impactful Individual
Thank you for your links. However when reading through them, I don't see something mentioning that when you import data: "So when you are using Import mode for your tables, whenever the report is used, the dataset is loaded into the capacity's memory".
It only mentions that when you publish a report, the compression ratio's vary when available in Service. That during the publish/ refresh state the VertiPaq storage engine is deals with an increased memory usage, but nothing is mentioned when an end user requests a dashboard of the data model.