-
We are using Janusgraph with Hbase backend for storing large data lineage graphs. The basic usage is to find a node, then do an impact analysis by traversing all the nodes that are affected by this node (recursively). The speed I am currently getting is about 620 edge traversals per second. I consider that quite slow. Here is the gremlin query:
Here is our configuration/environment: Janusgraph server 0.6.2 RAM 32 GB 13 nodes Is this speed normal? Is there way to make the query run faster? Would Cassandra be better for our usecase? I am adding all the details - explain step output and configuration: Query explain output:
gremlin-server.yaml:
janusgraph-hbase.properties:
Traversal Metrics:
|
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 23 replies
-
Can you show the profile result by adding You mentioned 13 nodes - how many of them are running JanusGraph server and how many of them are running HBase? |
Beta Was this translation helpful? Give feedback.
-
Just to be sure that "storage.batch-loading=true" is not interfering with the query.batch setting. After initial loading of the graph, the proper setting is "storage.batch-loading=false". |
Beta Was this translation helpful? Give feedback.
Just to be sure that "storage.batch-loading=true" is not interfering with the query.batch setting. After initial loading of the graph, the proper setting is "storage.batch-loading=false".