I am trying to set a benchmark for when we will create QVD files rather than loading for each model. Does anyone have some pointers for me? As an example if there are more than 500,000 records and these are going to used in more than one model then use a QVD. I am trying to standardise how we are doing things but just wanting to know if anyone else has any suggestions re the number of records.
If you are going to implement QVD load in a file, I find it best to implement across all data imports(except inline). If QVD load is used consistently, it helps organize and simplify the load process. The question of memory usage is another beast, depends on # of records & # of columns.