Monday, July 5, 2010 at 11:00AM | Derek Stainer
Robert Scoble and Emil Eifrem discuss Neo4j, graph databases and BBQ restaurants.
Couple of notes taken from the interview:
- There are a lot of interesting integrations of social and geographic graphs that have yet to be explored.
- Neo4j more general purpose and supports infinite depth. This is contrast to a specialized solution like Twitter's FlockDB which is geared towards Twitter's unique needs.
- Graph databases excel whenever the value is in the connections.
- How scalable is Neo4j?
- It's really challenging to provide horizontally when dealing with a graph database.
- Neo4j scales to 12 billion nodes and relationships.
- Currently relies on replication of the entire graph across machines.
- In Neo4j 2.0 they are implementing transparent partitioning. Transparent partitioning essentially attempts to place isolated clusters of a graph, on their own machine. It's a difficult problem to solve simply because two isolated clusters can become connected with just one link.
- Neo4j uses AGPL license. Essentially if your service is open source, Neo4j is open source. If your service is going to be closed source you need to purchase a commercial license.