I have not used the others but I agree, ArangoDB should meet your needs. If you have worked with RDBMS and SQL before Arango will be a easy transition. AQL is simple yet powerful and deployment can be as small or large as you need. I love the fact that for my local development I can run it as docker container as part of my project and for production I can have multiple machines in a cluster. The project is also under active development and with the latest round of funding I feel comfortable that it will be around a while.

Hi Jaime. I've worked with Neo4j and ArangoDB for a few years and for me, I prefer to use ArangoDB because its query sintax (AQL) is easier. I've built a network topology with both databases and now ArangoDB is the databases for that network topology. Also, ArangoDB has ArangoML that maybe can help you with your recommendation algorithims.

Hi Jaime, I work with Arango for about 3 years quite a lot. Before I do some investigation and choose ArangoDB against Neo4j due to multi-type DB, speed, and also clustering (but we do not use it now). Now we have RMDB and Graph working together. As others said, AQL is quite easy, but u can use some of the drivers like Java Spring, that get you to another level.. If you prefer more copy-paste with little rework, perhaps Neo4j can do the job for you, because there is a bigger community around it.. But I have to solve some issues with the ArangoDB community and its also fast. So I will preffere ArangoDB... Btw, there is a super easy Foxx Microservice tool on Arango that can help you solve basic things faster than write down robust BackEnd.

It really depends on what you are trying to do.

Using the graphdb only as a data-store to support an app, and don't expect a complex graph? Go with ArangoDB. It's a bit lighter, and provides more flexibility in what you store in a node. However, with that flexibility comes a data-structure that is more of a mix of a graphdb and a non-relational db. It lets you do things that you really shouldn't do, like put nodes within nodes, instead of connecting them via an edge.

Need to be able to easily reflow your graph, analyze it, run algorithms against it, update it from an api, or efficiently query large and complex patterns? You'll need Neo4j.

The only drawback with Neo4j is a steeper learning curve.

In case you need only the functionality of GraphDB, operate with not too large graphs and used to Cypher query language then Neo4J is obvious choice, otherwise I would recommend ArangoDB. The latter has a hybrid storage model (Document, Key-value and Graph stores) and provides much more capabilities including the native deployment in DC/OS (ArangoDB was one of the first DBs deployable there) and Kubernetes. Also, you can deploy DApps and microservices using JS directly on top of ArangoDB, it's a full-stack solution. Neo4J is a more mature GraphDB, has more elaborated visual UI for graphs, wider community and slightly better documentation, though ArangoDB also provides pretty good documenntaion and direct support.

Having used both, I would definitely recommend ArangoDB. Not only is it multi-model, but it is more intuitive to use as a developer. And most importantly, it supports nested JSON in the graph nodes which in my opinion is a deal-breaker for neo4j. A storage engine should not influence the shape of your data models. A simple example is "node.location.lat", you can't do that with neo - you have to create a new 'location' node and connect it with an edge.