Hi Stéphane, To improve the writes you should use sharding, no replication: in fact, the more nodes you have, the more replicas OrientDB has to manage. With replicas of your database you improve reads.
In 2.0 we introduced the new "asynchronous" mode to go much faster on writes too. Lvc@ On 18 December 2014 at 12:51, Stéphane Schild <[email protected]> wrote: > > Hi, > > I have done some tests in a lab on write scalability in distributed mode. > > My lab consists of two Win7 machines (also tried on Linux with the same > results) linked by Ethernet network, that are in the same OrientDB cluster. > I have left the default configuration, except the write quorum that has > been set to 1 to see if it improved the results. > > The object are created using the Java Graph API. > > Test 1 (1 client, 1 node): > 10'000 vertices created. > Time: 00m50s > > Test 2 (2 clients, 1 node): > 20'000 vertices (10'000 per client) created. > Time: 01m00s > > Test 3 (4 clients, 2 nodes): > 40'000 vertices (10'000 per client, 2 clients connected on node1, 2 > clients connected on node2) created. > Time: 01m55s > > I have two surprises: > - Test1 and Test2 take approx the same time to run, which was not expected > - Test3 takes twice the time as Test2, which was not expected either. I > thought that adding a node would increase the global performance by almost > a factor of 2. > > Could you guide me if I'm doing something wrong ? How do you explain my > results ? > > Thanks > > -- > > --- > 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.
