better performance/best practice in DQ - added flags in fact table vs multiple filters in CALCULATE

[fsfs]

hi

the case is as follows:

- I have the facts that are binary in nature but aren't to be aggregated directly (due to business rules, as they are the same in granularity but different in interpretation) and in DAX they need to be put as not a very complex, but still robust set of CALCULATE modifiers (several direct filters + some summing of columns, maybe even using iterator with RELATED sometimes)

so I have two options:

1) go with a classic CALCULATE( aggregation, modifier1, modifier2, ... )

2) or to generate flag columns during import, that will reduce the DAX from the above example into SUM(single column) - or worst case scenario CALCULATE( aggregation, flag = something)

I expect to generate a set of up to 5 flag columns with 0/1/NULL

I estimate the fact table to grow into 100 million rows within a year

what will be a better performing option here using Direct Query mode?

[rbriga]

it possible to generate those flags at the source?

For example, through a SQL view?

I believe that would be the best performance option. 

[fsfs]

yes, that's the only option we are taking into consideration

either normal model with more complicated measures

or additional flag columns in an SQL query during import