Since you are using HBase (open source/Cloud db), I do not see the reasoning of using oracle (proprietary / RDBMS). Its sort of one step back in my view. I think irrespective of complexity, scale of requirements all analytic data can be gathered by using a set of MapReduce jobs on Hadoop with help of HBase coprocessors and this data can be stored on hbase itself. For faster reads you can use a caching front end tool such as Redis / memcached.
For realtime analytic data you can use Hadoop pipeline capability.
Hope this helps. We have been using this combination since a while and the results are astounding.