The world we live in is defined not just by the things that are around us but also how they are connected and influence each other. The same goes to the data. The business value that the data we have at hand is defined not just numbers that represent them but the way these numbers are connected to each other and the way they influence each other. Traditional databases models like RDBMS are really stable and have stood the test of time. But like all human inventions they have their own inherent defects. The way the data is stored in RDBMS is not really modeled after the real world meaning there is limit to the amount of complexity that we can get with these systems. Beyond which it becomes really hard and lots of energy is spent on just running the system rather than the solving the problem at hand. There are many new solutions out there that promises this solve the problem. But graphdb is one of the promising solutions out there.
To put in simple words the data in graph database is stored in nodes and relation between them rather than tables thus it is more natural way of representing data than in tables. To represent complex relations you don't need to have many tables connected in complex way thus increasing the performance. such systems are called as GDBMS (graph database management system).
Why would want to have your business data in graphdb?
RDBMS is generally good for transactional data since in general erp systems the transactions have to deal with only particular part and mostly they don't care about complex relations but business intelligence is all about finding existing relations and predicting new ones. So graphdb is instantly the beset way to store data in business warehouse.
Can existing RDBMS system be used for graphdb storage?
There are many solutions out there in the market that provide implementation of graphdb on existing RDBMS system thus reducing the need for extra investment for such systems by simply providing an abstraction layer over an existing RDBMS so that the end user gets the best of both worlds the graphdb and RDBMS. The systems like hana which boasts for its in-memory processing capability can be made use much more efficiently if some such abstraction layer is implemented on top of it so that the users who need such systems can use the abstraction layer while others can choose to ignore it.
Completely new graphdb systems have some problems but hybrid systems don't have them and they are bound to succeed and it may be the path that business intelligence might pursue in future to drive business forward.