Hi Pablo,
If you really understand Data Warehouse, you'll see that part of the problem is to find out what is the data you are going to need. The fact is that there's no such thing as a Data Warehouse model for dotLRN. It depends ont he data you want to gatherm and what kind of relations you want to monitor.
The first task is to understand the relations under OpenACS, I mean, how the objects are related to each other. You will realize that we have a great advantage: the metadata system. You'll see that there's a lot of data you can gather, and use this to build up your datamarts.
After that, you can create your Multidimensional model, considering whatever technology you want to use. OpenACS basically uses a relational database, based on objects. A good tip for dimensions is to choose wich objects you want to measure, I mean, wich ones you are going to see as a dimension, and then build your multidimensional model.
If you get into this part, building the cubs (cubos) can be very easy: you only have to know how you want to see the relations with each other.
As you have the data you need, just choose the ETL tools of your preference to extract the data. Pentaho has a very good suit for BI (www.pentaho.org), wich is basically Open Source, but you can choose whatever you want, even SQL scripts, to extract the data.
This subject is slightly related to my research right now, and let me know if I can hep you somehow.
Best regards
