creating a copy of the OLTP database is good choice
you can use the log shipping function to do this (or other methods)
(how do you plan to synchronize your databases?)

generating cubes will not lock your database if you do incremental updates
on a  optimized cube.  (or just a litlle lock)
Partitions will help you for the cube process step.

But your design appear to be good.

garbage in garbage out = bad data quality in the source = bad quality in
output (non existant clientID, duplicated product name, duplicated
records...). so an ETL tool can improove the data quality during the
process. or if you want to synchronize multiple sources (like reference
tables and operationnal tables)
But these steps reduce the loading time by adding data cleansing time.

if you think next step (which is SQL 2005) you'll have a lot of new features
to help you.
for example, you will be able to load directly  a partition of your cube
while you'll load your database, so 1 read and 2 destinations
(and there is a lot of options)