Graph DBs vs. Document DBs vs. Triplestores

Question

This is a somewhat abstract and general question. I\'m interested in the inherent (as well as implementation-specific) properties of different approaches to persist unstruct

Answer 1

Just a small correction to amk answer: Tinkerpop also contains an adaptor for ArangoDB, see https://github.com/triAGENS/blueprints-arangodb-graph/wiki/Gremlin. So you can use Gremlin queries with ArangoDB.

In general, multi-model databases like ArangoDB or OrientDB allow you to use all the nice features of document databases (schema-free, indexes) together with graph structures. A vertex or an edge is simply a document as in a document database. You can have as many properties or even embedded documents as you like. You can define hash, range, fulltext, or geo indexes on these documents. Or you can forget about the document structure and view your documents as vertices and edges, using Gremlin or some traversal language to investige the underlying graph.

As for the question "are we doomed with polyglot persistence": Independent of the document / graph database question, I belive that RDBMS will be around a little while longer. So, the answer to that question is: "yes, that is very likely".