August 17, 2018 at 11:18 am
I say this because accumulation in that way you and others explain it is a farce, it's not accumulating the records,
Hardly a farce. Go back and read your Kimball on the three types of fact tables. Two of them are transactional in nature and generally only involve adding new transactions. These transaction rarely, if ever, require correction. I've done several fact tables of this type, too. They are easier, but the transactions are complete out of the gate. One of them is an accumulating snapshot--the tracking of some object through its life. These fact tables usually has a lot of different dates associated with it.
The updating doesn't interfere with the reports. Processing a dimension using an update rather than full leaves the cube intact. You could also process the entire database, which doesn't get committed until completion. There is a few second window of unavailability, but in practice this has not been an issue.
Viewing post 31 (of 30 total)
You must be logged in to reply to this topic. Login to reply