To many onlookers, it may seem that the NoSQL-hype is solely focused on scaling data. Many NoSQL databases are designed such that they can horizontally-scale with relatively ease. This is accomplished by making use of data structures that are optimized for sharding. Such data have limited to no direct references between each other. Therefore, the problem of referential integrity does not exist and data can be massively parallelized. Examples of such systems include Amazon’s Dynamo, Google’s Big Table, Apache’s CouchDB, and so on.

In stark contrast to this design choice, on the other side of the NoSQL spectrum, there exists another design choice---the highly interconnected, direct referent data structure of the graph database. Graph databases allow users to solve problems by moving through their data in intelligent/directed ways and with an arbitrary depth. This style of data processing is known as the graph traversal pattern. This pattern is difficult to efficiently achieve with systems that only allow for the joining of data through the use of global indices. The graph traversal pattern is all about local, index-free traversals.