storage.parallel-backend-ops
Hello i am a bit confused wth the docu. Do i enable or disable storage.parallel-backend-ops? I have disabled the cql executor and configured my connections/requests but i don't know about parallel-backend-ops and also how to configure gremlin gremlinPool/threadPool accordingly?
1 Reply
Hello, I have shorter and longer answers to your question.
Shorter answer: Soon (starting from JanusGraph 1.0.0) this pool won’t be used. Moreover in the latter JanusGraph version this pool is automatically disabled if the underlying storage backend supports multi-key operations. However you need to use it for 0.6.x JanusGraph versions.
Longer answer:
This pool is responsible for parallel execution of multi queries which are used as part of the JanusGraph optimisation.
If you use JanusGraph 0.6.3 - you need to enable it and setup pool size in accordance with your server resources and usage pattern (i.e. it’s good to use big pool size like 1000 - 3000 to improve parallelism).
However, if you are using the unreleased version of JanusGraph 1.0.0 (rc-2 for example) then this pool is not used anymore for storage backends which support multi-key operations.
HBase supports multi-key operations in 0.6.x JanusGraph version.
CQL supports multi-key operations starting from JanusGraph 1.0.0 (instead of a thread pool the usage of async driver capabilities are leveraged).
In the later (yet unreleased version of JanusGraph) multi-query optimisations were added for most of the steps which should improve performance for many use-cases.