Yup, that is our assessment, too. On 1-3 joins there is a chance to optimize the queries in an RDBMS. But still you have indexes instead of references, and the overhead of serialization through the JDBC driver. Anyway, it even shows that Java based DBs are not necessarily slower than C based.
Then, as you mention, this is the simplest example of a graph traversal, since it is a straightforward join for every depth, In reality, things are more complicated than so. Traversal speed in this example is about constant at 600.000 returned nodes per second for Markos MacBookPro, which might be a bit slow, but it is - constant. Depth 5 took 30s for 17M returned vertices. Marko, would be interesting to se a diagram showing the MySQL and Neo4j mean times for different depths plotted in one graph? MySQL degrades as the data volume increases, and as the depth increases (at depth 5 MySQL just hangs). Cheers, /peter neubauer COO and Sales, Neo Technology 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://www.tinkerpop.com - Processing for Internet-scale graphs. http://www.thoughtmade.com - Scandinavias coolest Bring-a-Thing party. On Tue, Mar 30, 2010 at 5:02 AM, Laurent Laborde <[email protected]> wrote: > Hummmm... I'm surprised. > I expected the test to be unfair in favor of mysql for a simple graph > traversal. > In fact, i still think that with good mysql configuration and good sql > tuning, mysql can be faster. > > Of course, if you add different kind of relationship, node property, > relationship property and complex query, i'm pretty sure that neo4j > can be much much faster than any rdbms for querying graph based data > :) > > (just think about the how many nested sql join you would need if you > havenode/graph properties and many differents kind of relationship (eg > : social network)) > > -- > Ker2x > > > On Tue, Mar 30, 2010 at 1:37 AM, Marko Rodriguez <[email protected]> wrote: >> Hi guys, >> >> Sorry if you have already seen this (Twitter, Gremlin mailing list, etc.), >> but for those that haven't, this is of potential interest to all you Neo4j >> users. >> >> http://markorodriguez.com/Blarko/Entries/2010/3/29_MySQL_vs._Neo4j_on_a_Large-Scale_Graph_Traversal.html >> >> Take care, >> Marko. >> >> http://markorodriguez.com >> _______________________________________________ >> Neo mailing list >> [email protected] >> https://lists.neo4j.org/mailman/listinfo/user >> > > > > -- > Laurent "ker2x" Laborde > Sysadmin & DBA at http://www.over-blog.com/ > _______________________________________________ > Neo mailing list > [email protected] > https://lists.neo4j.org/mailman/listinfo/user > _______________________________________________ Neo mailing list [email protected] https://lists.neo4j.org/mailman/listinfo/user
