I've not been able to understand the role of a 'connection' in the execution of a query on a host.
cass_cluster_set_core_connections_per_host sets the number of connections made to each server in each IO thread. Why is this parameter configurable? What's the benefit of having more than 1 connection per IO thread? What decides the optimum value for this parameter?
This function sets the initial number of connections per host, and it could be increased up to value set by cass_cluster_set_max_connections_per_host function (by default it's 2). New connection is created when the number of in-flight request is greater than set by cass_cluster_set_max_concurrent_requests_threshold (default 100). You can set higher value with cass_cluster_set_core_connections_per_host if you know that you'll generate the big number of in-flight request from the beginning - in this case you won't spend time opening an additional connection when you're executing requests.