If I don't batch, I usually hit a limit of somewhere in the vicinity of ~500 statements/second on a single threaded client using the Cypher HTTP endpoint. With batching on the transactional Cypher endpoint it's way faster, and should handily beat your 8 second time. So, no, it's not generally slow. Disk latency seems to have a big effect when you do small transaction creates over REST/HTTP as well, but for it to take more than an hour, something else has to be at play--that's like half a second per node (assuming ~10k nodes). Is your neo4django client running over a slow network to the db or something?
Wes On Fri, Feb 7, 2014 at 3:45 AM, Antonio Mignolli <[email protected]>wrote: > Hi, > I know there's some debate about real performance with neo4j. > > I created the almost complete set of italians regions, cities and > towns, in a simple graph > where each Region has Provinces, each Province has Districts. > We have a total of about 8000 districts, so this set of data is not so big. > > With Neo4j graph creation it tooks more than a hour on a Core I7 with > 8GB of RAM. > Trying the same on a SQL database, less than 8 seconds. > > I'm not creating cypher queries directly, I'm using Django for both > examples. > So, neo4django when dealing with neo4j. > One can think it's all neo4django overhead, but I seriously doubt it. > > Is neo4j generally slow on create nodes? > > -- You received this message because you are subscribed to the Google Groups "Neo4j" 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/groups/opt_out.
