I think, it depends on customers in database, currently, it is 2.7M customers.
Either use RDF and SPARQL and leverage your graph potential or use relational on HBASE, the current solution does not seem to work for you.
On the graphing side, learn to use R. It will serve you a lot better than trying to make graphs with excel.
27 billion triples can easily be worked with using e.g. Virtuoso 7.2 on commodity hardware. e.g. 256gb RAM+ 2Tb of consumer SSDs.
Your current setup does not leverage either the strengths of RDF or of HBASE. Moving of HBASE and onto a SPARQL/RDF store will be less work and I expect to have a higher ROI.
Remember SPARQL stores have improved by leaps and bounds in the last few years so the home grown HBASE solution is no longer state of the art.