I setup visualvm with a jmx connection to the server running neo4j I wasn't able to CPU profiling working, it had an error message saying getThreadCpuTime was expecting long but got something else. I do have thread and heap dumps though if those would be useful.
When I was running the batch upload I saw that the CPU usage never went over one core ~15% total cpu usage. Also I noticed that when I did a batch job of 1.5k nodes, 3k relationships the network usage would spike and be consistent and then fall for the upload and then spike and fall again for the read back from the server. However when I did 15k nodes and 30k relationships the network would spike, fall and then have short regular spike of only a couple KB/s. Is there some internal paging going on with the REST server, why doesn't it transfer the file all at once and then process it? The 1.5k job took about 25 seconds, the 15k job took 1220 seconds. Below is the configuration I have for the server, the machine itself has 32GB of ram and 8 64bit cores: -XX:+DisableExplicitGC -XX:+UseParallelGC -XX:+UseNUMA -Xss1024m -Xms2048m -Xmx4056m Neo4j properties: neostore.nodestore.db.mapped_memory=300M neostore.relationshipstore.db.mapped_memory=600M neostore.propertystore.db.mapped_memory=600M neostore.propertystore.db.strings.mapped_memory=1024M neostore.propertystore.db.arrays.mapped_memory=1024M cache_type=strong use_memory_mapped_buffers=false -- View this message in context: http://neo4j-community-discussions.438527.n3.nabble.com/Neo4j-Batch-uploading-to-REST-server-tp3477770p3510431.html Sent from the Neo4j Community Discussions mailing list archive at Nabble.com. _______________________________________________ Neo4j mailing list User@lists.neo4j.org https://lists.neo4j.org/mailman/listinfo/user