Comment by girvo | Hacker News Reader

Comment by girvo | Hacker News Reader

Parent Full thread

I read through the OP, and it's convinced me to at least have a look at OrientDB. It seems interesting, though I've never used a Graph database before, let alone something like this.

I'll be honest: I've pretty much only ever used Relational databases. I've not even properly tried using Document databases (the hype just didn't overcome my laziness and inertia!). On big big projects, I just wrote the queries I needed, with a schema I defined, and let my DBA deal with the nitty gritty of it... I really should learn more about the various classes of data stores. I know enough to understand a lot of the trade offs inherent to the major types (hence why I've stuck with relational, it fits the way i think about application data, and I haven't needed Web Scale(tm) for those projects, so horizontal scaling wasn't ever an issue), so I'm really curious how this one holds up under close scrutiny, and whether I can map how I think about data to how Orient works (or whether it's way of tackling it is better enough than how I think to just embrace it's style entirely).

Anyone here have experience with it? What use case, data model, etc. what trade offs did you make, and why?

We are using OrientDB to store visitor stats on meople.net behind the scenes: http://meople.net/visitor/index.jsp

Document oriented tables are used. I like no-sql access to indexes:

  OIndex<?> index = db.getMetadata().getIndexManager().
      getClassIndex(klassName, indexName);
  Collection<ODocument> = index.getEntriesBetween(start, end, true);

Nothing to say here: it just works.

I'm in the same boat. I've been looking at Neo4J, Titan, etc., as part of research for a new application I'm building. Our data set is fairly tiny (<<10M entities, almost certainly), but the information we're tracking is highly graph-oriented.

I've been trying to decide between a graph db, which seems like a better fit for the real world relationships but is entirely new to me, and postgresql with exports to graph structures for analysis as needed.

In particular, I'd be interested in deployment experiences with Orient and other graph dbs. I don't have a sys admin and I'm not one myself, so a PaaS platform is highly attractive.

Neo4J has a Heroku "try" option, but there's no clear upgrade path. Titan is flexible in terms of backing store (Cassandra, PersistIt, etc.), but seems to require self-admin.

Anyone deployed and maintained OrientDB or any of these other tools?

In your research, have you found any good graph database not written in Java?

Triple/Quad stores (specialisation of a Graph DB for storing RDF) are available written in C (4Store, Virtuoso) as well as Lisp (AllegroGraph).

They support a standardised query language called SPARQL which is based on matching graph patterns.

ArangoDB (C/C++, http://www.arangodb.org/)

Here you have 15 Graph databases compared: https://docs.google.com/spreadsheet/ccc?key=0AlHPKx74VyC5dER...

I'm building the app in Java, so I specifically looked for those. Almost all of them have REST built-in (or easily available via the stack at http://www.tinkerpop.com), so the client could be language-agnostic. Look at http://bulbflow.com for Python, for instance.

http://www.sparsity-technologies.com/dex.php

Heroku will have another Neo4j option in January that will offer an upgrade path.

Disclaimer: I work for Heroku on Add-ons.

I've used it in my Java Swing GUI app which I open sourced (https://github.com/jjk3/scrape-it-screen-scraper) in year 2012, because the other Object database cost licensing money (in small letters).

It was the only joy I got from working with a Java Swing. It's excellent to just store Java Objects directly to the local filesystem. If you don't like that, you can setup a OrientDB server. The writer of OrientDB was super helpful too (always on google groups helping out). It's flexibility outdoes MongoDB.

I'm surprised that it's catching on now...after 2 years.

Reply on news.ycombinator.com