For certain reasons I am using AnyScale to parallelize computation of nodes and edges to insert into my Neo4j DB.
I saw on the Neo4j website that they recommend having only one driver open for each application since due to it having to set up a connection pool every time and having multiple drivers connected at the same time can be costly.
Is there any significant performance downside to having each of nodes owning it's own driver connection or should I apply something like MapReduce to ensure that a minimal number of nodes can connect to the Neo4j db?
Asked by Kevin Y.
(1 rep)
Oct 31, 2024, 04:00 PM