Anybody with experience in using graph db with complex traversal? AND use it from Erlang? Advantages? Small print?
It seems that default choice is neo4j for some reason.
I like Phoebus (distributed graph querying in Erlang, based on Google's Pregel algorithm), but it seems to be abandonware..
Have a look at OrientDB which is a good alternative to Neo4J. OrientDB is fully open-source, contrary to Neo4J which has a commercial license. Also, OrientDB supports the standard blueprints API and is based on RB-Tree algorithms that are very well performing for various graph use-cases.
What most of the graph databases have in common is that they scale up only. They can scale-out, although as read-only instances in general. This means that if your amount of data is excessive you will need a very large server as well ( memory and storage ).
A distributed graph alternative is Titan. Titan is a distributed graph database, which I believe is based on Hadoop. Titan is designed to scale-out and can offer an interesting approach for massive paralelism, with some overhead. There are use cases where this is more appropriate, such as similar to Google Pregel use cases.
you can represent a graph in an RDF database like Stardog. There should be some off the shelf code out there for doing SPARQL queries using erlang.
Wikipedia has a pretty nice list of triple stores.
