Multiple Facts in single subject area

Question

Hi everyone,
Is there an optimal way to get multiple facts into a single subject area? The only way I've managed to get multiple facts into a single subject area is to go the route of creating hierarchies for all dimensions, setting up all the content levels for both the fact and dimension tables and finally creating aggregated measures for all the fact tables. However, when you setup these different hierarchies and content levels, the queries which are generated tend to get fairly nasty and seem to be slow. Another pesky issue is that I've had to create pseudo columns for all facts, so that if a report is using one dimension from FACT A and one dimension from FACT B, then you have to include this pseudo column from both facts so that OBIEE knows which tables are involved in the query. Without including these pseudo columns in the report, the OBIEE engine gets a little confused and puts up an error.
I'd prefer to find a better way to do the multiple fact implementation without having to have the massive queries and pseudo columns, but I'm not sure what the other options out there are. Anyone know the best way to accomplish this?
-Joe

Answer 1

Hi Wildmight,
You are exactly right.
Here's an example of what I'm trying to accomplish. We have a TIME SHEET fact table which has things like a work date, an employee ID, a work order number and the hours which were worked. The work orders can be against one or more pieces of equipment. There is a second fact table which models the one to many relationship between work orders and equipment.
The model would look a little like this:
Join on WORK ORDER NUMBER Join on WORK ORDER NUMBER
TIME SHEET FACT <-----------------------------------------------WORK ORDER Dimension ----------------------------------------------> WORK ORDER EQUIPMENT FACT
We would like to create a report that shows the number of hours from the time sheets fact grouped by the piece of equipment from the work order equipment fact. I've been able to setup this report using the method I described in my first post, but like I said, the generated SQL is massive and seems somewhat unnecessary.
I think I'd have to say the grains of the facts are different since they providing to different pieces of information.
I haven't done too much multiple LTS experimentation. I'm working from a single data warehouse source which should be robust enough to handle any reporting requirements. Even though I know OBIEE can do some clever combining of tables using the LTS; if data sets need to be consolidated, I'm trying to keep that work in the database via ETLs in order to keep the work that OBIEE has to do down to a minimum.
-Joe