Hi Jemma,
I think you should approach your problem in several ways.
1. I think the reason that the load is slow is not because of CUBE modeling, l would think that the slow load is the symptom of information being stored in the PSA, most likely there are millions of records. This is not avisable.
a) I'm not sure why it is stored all information in PSA since you have it in the CUBE.
You can delete requests in PSA and mantain only keep the 3 past months and I can assure you the load will be within 1 hr.
Additional Comment: Please check with your basis team and check for the largest tables in your
system. If the biggest are PSA or something like /BIC/000003154, you must take action
DELETING all your PSA's of your biggest models. Create a PSA deletion step in Process Chains.
2) TIP: I would strongly reccomend to create a copy of your cube and mantain it intact with all information.Connect ORIGINAL CUBE -> COPY CUBE with transformation and do loads per month. YOU NEVER KNOW what can go bad.
IF you have a MULTIPROVIDER on the TOP of your CUBE, you can mantain ORIGINAL CUBE as the loading and COPY CUBE AS HISTORIC, THIS also will load much faster and should not impact queries or loads.
b) If you need to re model this cube, you can use function module with Tx se37 RSDU_INFOCUBE_TABLE_SIZES to check for the CUBE and the dimension tables SIZE so you can model it in a better way.
after remodeling, and having teh COPY CUBE as backup, you can reload all PSA.
AFTER you finish reloading your cube,
I would reccomend you to migrate from 3.x to 7. Please check this link as a starting point.
As you can see there are several points to consider in your issue.
Step by step you can solve it succesfully. Always take precautions.
Best Regards.
Hope it helps. you can contact me directly through gjunoe@hotmail.com
Regars from Mx.