Hi Dimitri, Also, Blueprints assumes "auto transactions" (that is, every delete/write is a committed transaction). You can turn that feature off if you want better write/delete performance.
One of my next big goals with Blueprints is a general-purpose transaction model so that all transaction-based graphs have the same interface through Blueprints. No "graph instanceof NeoGraph" needed. Hope that helps, Marko. http://markorodriguez.com http://tinkerpop.com On Sep 10, 2010, at 4:28 PM, Dmitrii Dimandt wrote: > > Yeah, I guess that could be a problem (or the problem). > > Hmmm... I'll try and run the "benchmark" on a less busier machine and see > what happens then. > > Thanks for the tip! > >> FWIW, I've noticed that neo4j seems to be fairly sensitive to disk I/O >> performance. >> >> I couldn't get it to work at all on an NFS mountpoint. >> >> It works much better on a Solaris tmpfs partition instead. >> >> >> >>> I got cali's benchmark result. Well, sort of benchmarks: >>> http://github.com/dmitriid/cali/wiki/Benchmarks-of-sorts There's neo4j, >>> TinkerGraph and OrientDB >>> >>> For some odd reason neo4j performs abysmally, though yesterday it was way, >>> way better: http://twitter.com/dmitriid/status/23912123904 >>> >>> Will have to investigate further. >>> >>> >>>> Hey, >>>> >>>>> Yeah, it'll be my next step, I guess :) Blueprints make this stuff >>>>> super-easy :) >>>> >>>> Yea---write to the Blueprints interfaces and then you can plug and play >>>> your graph backends. >>>> >>>>> I also see plain simple FS in there as well ;) >>>> >>>> Ha. Yea---I started that, but never went far with it. In short, it makes >>>> your file system look like a property graph (e.g. subdirectory, >>>> contains, etc.). When Gremlin moves into more advanced I/O support, I >>>> think I'll revisit FileGraph as that might be a good way for Gremlin to >>>> interface with the operating system's file system. >>>> >>>>> BTW, I've also noticed there's mongodb support for storage (?) How >>>>> would you use that? >>>> >>>> Yea. So, long long ago, Peter and I wanted to make a single interface >>>> for all NoSQL systems. The JSON databases of CouchDB and MongoDB were >>>> the next up after property graphs... Currently, in Blueprints, there is >>>> support for MongoDB and TinkerDoc as JSON databases, but its not fully >>>> tested, not fully thought out, etc. etc... Hopefully in the near future, >>>> we can get some steam to flesh all that out better. >>>> >>>> Good luck with Cali, >>>> Marko. >>>> >>>> http://markorodriguez.com >>>> http://tinkerpop.com >>> >>> _______________________________________________ >>> Neo4j mailing list >>> [email protected] >>> https://lists.neo4j.org/mailman/listinfo/user >>> >> >> >> -- >> Rick Otten >> [email protected] >> O=='=+ >> >> > _______________________________________________ Neo4j mailing list [email protected] https://lists.neo4j.org/mailman/listinfo/user
