In my use case neo4j outperformed MySQL by a factor of 377 ! That is more than two magnitudes). As known one part of my PhD thesis is to create a social newsstream application around my social networking site metalcon.de. It is very obvious that a graph structure for social newsstreams are very natural: You go to a user. Travers to all his friends or objects of interest and then traverse one step deeper to the newly created content items. A problem with this kind of application is the sorting by Time or relvance of the content items. But before I discuss those problems I just want to present another comparission between MySQL and neo4j.
via NoSQL databases