Hi Stefan, The reason why I don't want to feed this topic are pretty much the same as I wrote months ago. In few words:
- The benchmark has been created by a vendor to exalt its specific features. - Who wrote the benchmark start publishing results before asking to vendors to tune each products. This was obviously a great marketing operation and seems it worked quite good for a small vendor with poor attention from media before that benchmark. - Most of the tests stress more the Node.js driver than the server. When I run them the CPU taken by OrientDB server was 0,1% and 99,9% was the node process! Unfortunately we don't have a super fast Node.js driver, even if Emanuele yesterday pushed the first version of the hybrid Node.js/C++ and results are already very positive. Soon he will release some update on it. - I'm not 100% sure, but I remember to read that they published the average of 5 runs. The bad is that the server was restarted every time. This is against any real life usage, where the server stays up & running and starts to optimize things as soon as you use it. This means also that all the optimizations that Java HotSpot does by compiling under the hood the most used part, it's completely unused. - For the reason above, you should write a warmup phase at the beginning that merely loads the entire database in RAM, and nobody spent too much time to write a proper warmup for that benchmark. - With a small dataset that fits in memory, the advantage of the index-free adjacency are zero. This dramatically changes when you have huge datasets that don't fit in RAM. Then you appreciate the O(1) speed of loading edges against O(logN) of using an index. That said, I'm not saying OrientDB is the fastest DBMS in any single use cases. We already know the bottleneck of our architecture and we have a clear vision and a roadmap about how to address each one. Best Regards, Luca Garulli Founder & CEO OrientDB <http://orientdb.com/> On 22 October 2015 at 10:15, <[email protected]> wrote: > Hi, > > This is now the fourth day that this goes undisputed and I find that odd. > > On behalf of happy OrientDB users I want to request some information here > regarding this. > > > 1. Is this test conducted fairly? > - To my surprise I see the js driver is being used (for one) > > 2. Does it accurately portray any speed difference > - Compared to the first test that can hardly be > > 3. Why is this not addressed at all here? > > In some cases things are best left ignored. I don't believe this to be one > of those cases. > > Regards, > -Stefan > > > > > On Tuesday, 20 October 2015 22:21:44 UTC, [email protected] wrote: >> >> >> ok, this ArangoDB test frustrates the hell out of me. >> >> I know that their previous tests have been disputed but I hope this will >> be addressed as well. >> >> best regards, >> -Stefan >> >> >> On Sunday, 18 October 2015 21:38:43 UTC, [email protected] wrote: >>> >>> Hi, >>> >>> Has this been addressed at all? >>> >>> >>> https://www.arangodb.com/2015/10/benchmark-postgresql-mongodb-arangodb/#more-8721 >>> >>> Regardsm >>> -Stefan >>> >> -- > > --- > You received this message because you are subscribed to the Google Groups > "OrientDB" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/d/optout. > -- --- You received this message because you are subscribed to the Google Groups "OrientDB" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
