@Hubert Dudek​, creating a managed table means persisting the data frame (writing the content of the dataframe to storage) .

Imagine you have a Spark table (in delta lake format) containing your raw data. Every 5 Minutes there is new data appended to that raw table.

I want to refine the raw data: filter it, select just specific columns and convert some columns to a better data type (e.g string to date). With VIEWSs you can apply this transformations virtually. Every time you access the view, the data are transformed at access time - but you always get the current data from the underlying table.

When creating a managed table you basically create a copy of the content of the raw table with the transformations applied. Persisting data comes at performance and storage costs.

Moreover, every time, I want to access my "clean" version of the data I have to specify the transformation logic again. VIEWs allow me to just access my transformed raw data without the need of manual refresh or persistence.