I want to make report from a infocube A with 100 million records. They are about the outfit sales records from 2011 to 2013. It is too slow to use the data in cube A directly. So I try to improve it with two different ways.
The first way:
I plane to create twelve infocubes with six dimensions , twelve characteristics and three keyfigures. They store one month sales data of each year in the time range respectively.(Totally three month's sales data in each cube)
Although the cube only has twelve characteristics, so I can put them into twelve dimensions and set the dimensions as a line item, am I right?
Then I should set the error handling as deactivated in the dtp's updated tab. This can improve the transfer performance. Anyother operation can I do to improve it?
In this case, I don't have any routine and transfermation on data format. That's all.
The second way
This way is so simple. I will create twelve aggregate on the big old cube and respectively restrict them on year characteristic. Make the same effect like logical partition on time. Can it work?