Neo4j 1.7 Released
blog.neo4j.org
blog.neo4j.org
-Andreas (part of the Neo4j community team)
- James, author of Bulbs (http://bulbflow.com), one of the Python bindings
I am aware of the Heroku addon, but what happens if you need more than that?
For write scaling, you would have to stay tuned. There's a team sequestered in England feverishly working on a sharding solution to spread write load and data set size across multiple machines without compromising query performance.
Heroku is just the first cloud service to offer Neo4j. We've been chatting with other PaaS providers, and some are moving forward with their own plans. A few enterprise customers simply orchestrate their own clusters (up to global, multi-continent spanning deployments) using EC2.
You're right that all of the NOSQL solutions don't have as many experts and professional services as the relational db options. Seems like a great opportunity. ;)
Also ThoughtWorks is using it and writing about it: http://www.markhneedham.com/blog/2012/04/20/neo4j-searching-...
How do you effectively use SemWeb tech? Obviously you can't build your whole system with a triple store backend; that would be too slow. Especially if you have a data explosion of millions and millions of facts.
I've thought of a neat little distributed p2p research project that exchanges news items like HN/Reddit with your interest graph as determined by the nodes you share upvotes with (or weighed by some scheme you can tweak). Do you simply use RDF/OWL ontologies like FOAF, SIOC, etc. as the data transport format and then actually store information in relational databases internally?
I want to do this so people can curate their own interest graphs, essentially; it's like prismatic, but software under your own control. Such a platform would be extensible since you can add new ontologies to support new needs. The Slashdot crowd, for instance, might moderate with tags such as 'insightful' rather than simply 'upvoting'.
And can you use these graph databases like Prolog systems? And can we make that intuitive for the end user?
Or as a more practical example, what about a molecule database? I'd need to store representations, properties, etc. There's a potentially limitless amount of factual information that I could store on one particular molecule. How do you make your software flexible enough to "know" what data you have on any given molecule and decide which "context" to portray it? (ie. if it's of medicinal use, synthetic chemistry use, is simply a common reagent, etc.) Is some of this just inherently kept in a relational database for fast querying with the graph database backing an asynchronously updating view?
The RDF/OWL approach has a nice advantage with a built-in notion of logical relationships between things, while being a bit more verbose in storage. Ontologies are a mixed blessing.
For the flexibility you mention, it's common with Neo4j to "organically grow" your data by simply elaborating the structures and relationships. This allows, for instance, multiple domains to be intermingled across a common base set of data, differentiated by the relationships they establish to the common data.
Cheers, Andreas