I recently joined a new team in my organization where they are reporting on replica of OLTP database which loads data from actual OLTP every night into the replica.
There are 25 reports which are complex in terms of number of joins but most of them are kind of detailed reports(with more text than figures) in the sense not aggregated. The reason given from the developers for not using DMR in FM is that there is no scope for such design keeping in mind the type of reports they have(not fact oriented).
For me, this design doesn't seem alright(firstly, reporting on OLTP though its a replica).
They are planning for a production move soon. So, I would like to know if anybody have worked on this type of model and any suggestions to make these reports work better in terms of performance or design(there isn't much time left but really wanted to try is possible).
-Sarah
Non-aggregated data as primary reporting output certainly benefits from a non-olap approach. Both DMR and traditional cubes are not the best way for reporting against such details, but you may consider adding additional indices to the replicated database for performance sake. The developers have a strong point here..