Microsoft Research has begun working on its own graph database, Trinity. Graph databases store data in terms of nodes and edges instead of rows and columns, making them quite effective for loosely and arbitrarily connected data. Potentially uses for this included social networks, movie recommendations, and related product searches.
Trinity is designed as a distributed hypergraph. Unlike a normal graph database that links pairs of nodes, each edge in a hypergraph can connect an arbitrary number of nodes. This can be used to pull together sets of information, such as a user’s profile or the contents of a web page. Both nodes and edges can store heterogeneous data in the form of key/value pairs.
For more information on large scale graph databases in production today look to Facebook’s distributed graph database, has been covered on InfoQ multiple occasions. There are also several commercial and open source projects including GraphDB, InfiniteGraph, HypergraphDB, and Neo4j.