Hello,

I'm using janusgraph/gremlin server for storing and searching graph data and the back-end is hbase.

In testing for the performance, i bursts the server with 100000 requests for searching for a particular node as the following query. This query is parameterized so my guess is sending a 100000 requests separately should not over load the server.

gremlin = "g.V().has(\"Instruction\",\"insId\",transInsId").out(\"Instruction to transaction\").has(\"customId\",customId)"

My graph is indexed for the label "Instruction" and parameter "insId" so the retrieval should be quick.


When i send the 100000 search requests average search time happened to be 40ms per search which is not acceptable for my requirement.

So what i did is increasing the number threads do the searching.
I divided the 100000 requests to 10 threads with 10000 per each and each thread created separate websocket connection to connect  and send the requests.
But i didn't get any increase in the overall speed.

Is there any settings in the janusgraph/gremlin server level to set?

Can you please give me a suggestion to increase the throughput of search requests?

Thanks
Dilan.