We could look at how much work would it be to use Neo4j as backend DB.
And if not much, we could try it and see if it would impact performance significantly.
(Still, optimizing data structure and queries is probably more promising.)

This would mean changing:

• The way we handle storage - maybe just a bit (Neo4j also just needs a directory)
• Property indexing
• Titan-specific APIs - triggers (listeners)
• what else?

(02:20:04) ozizka-FN: I have an idea
(02:20:16) ozizka-FN: we could write migration rules from Titan to Neo4j
(02:20:21) ozizka-FN: and then apply it to Windup.
(02:20:42) jsightler: We could probably use Neo4j as a backend, but I'm not sure that it would actually be better.
(02:21:26) ozizka-FN: http://db-engines.com/en/ranking/graph+dbms
(02:21:26) jbossbot: Title: DB-Engines Ranking - popularity ranking of graph DBMS
(02:22:07) ozizka-FN: Doesn't really say about quality, but at least about the probability of things getting fixed
(02:22:21) jsightler: It does?
(02:22:41) jsightler: Anyway, blueprints has a neo4j backend, doesn't it?
(02:34:53) jsightler: It does... it would be pretty easy to try it on neo4j if you wanted to.
(02:35:08) jsightler: (well, it wouldn't be that easy... maybe a few days work?)
(02:35:35) ozizka-FN: Some our features rely on Titan stuff
(02:35:42) jsightler: Right
(02:35:52) ozizka-FN: But could be worth trying WRT performance
(02:35:57) jsightler: The blueprints APIs are very leaky, so there would be some tweaking involved.
(02:36:22) jsightler: I'd be really surprised if the difference were significant, but it could be.