Marco, yes, small transaction require many disk flushes, so you are seeing the limits of the OS IO. However, we are working on a MUCH faster mode of relaxed writing, apart from minimizing the number of files to flush. We hope to have these optimizations as part of 1.4, which means they will appear hopefully in the next milestones.
HTH. Cheers, /peter neubauer GTalk: neubauer.peter Skype peter.neubauer Phone +46 704 106975 LinkedIn http://www.linkedin.com/in/neubauer Twitter http://twitter.com/peterneubauer http://www.neo4j.org - Your high performance graph database. http://startupbootcamp.org/ - Öresund - Innovation happens HERE. http://www.thoughtmade.com - Scandinavia's coolest Bring-a-Thing party. On Tue, May 31, 2011 at 5:45 PM, Marco Gerber <mger...@junisphere.net>wrote: > Hello everybody > > The performance issue results from a too small transaction size. If I > commit only after 1000 traversials, the execution time for every traversial > speeds up to ~ 300 times. The same thing happens on windows where the speed > is near infinity :-). It seems, that there are two problems with the current > implementation (also tested with 1.4.M03). One with the case of handling > transactions on linux itself and the other with handling small transactions > with good performance in general. > > jm2c, > Marco > > > -----Original Message----- > From: user-boun...@lists.neo4j.org on behalf of Marco Gerber > Sent: Tue 31.05.2011 08:58 > To: Neo4j user discussions > Subject: Re: [Neo4j] performance issues with ubuntu > > Hello everybody > > A collection of reference benchmarks would be great, but completely > independent of this, my problem is making my application on linux as > performant as on windows. > > Thanks guys, > Marco > > > -----Original Message----- > From: user-boun...@lists.neo4j.org on behalf of Rick Bullotta > Sent: Mon 30.05.2011 18:02 > To: j...@neotechnology.com; user@lists.neo4j.org > Subject: Re: [Neo4j] performance issues with ubuntu > > Hi, Jim. > > Not really thinking of benchmarks, which I agree are tricky to define and > even trickier to standardize. Plus, given the nearly infinitely cool things > you can do with neo, it borders on impossible. > > Rather, I'm just thinking of wikifying some of the platform specific best > practices and gotchas/known issues for each. > > Rick > > ----- Reply message ----- > From: "Jim Webber" <j...@neotechnology.com> > Date: Mon, May 30, 2011 11:55 am > Subject: [Neo4j] performance issues with ubuntu > To: "Neo4j user discussions" <user@lists.neo4j.org> > > Hi Rick, > > I concur that we should perhaps have some perf figures, but it's one of > those things that's easier said than done. > > For instance, right now we have performance tests running as part of the > Windows/Linux/Mac continuous build and we will fail the build if we drop on > those numbers. Yet creating representative benchmarks (similar to the TPC-X > benchmarks perhaps) is not only difficult, but will almost certainly be of > no use to users or customers when choosing a graph database or in designing > out their own solutions because your access patterns are likely to be so > different from the benchmark patterns. > > But if the community could come together and loosely agree upon some > representative benchmarks that would be useful to them > (platform/setup/access pattern) then we could get something into our build > and publishing numbers with a little effort. > > Jim > > > _______________________________________________ > Neo4j mailing list > User@lists.neo4j.org > https://lists.neo4j.org/mailman/listinfo/user > _______________________________________________ > Neo4j mailing list > User@lists.neo4j.org > https://lists.neo4j.org/mailman/listinfo/user > > > > _______________________________________________ > Neo4j mailing list > User@lists.neo4j.org > https://lists.neo4j.org/mailman/listinfo/user > > _______________________________________________ Neo4j mailing list User@lists.neo4j.org https://lists.neo4j.org/mailman/listinfo/user
