Hi Olga,
> Switching everybody to HiveServer2 has performance and scale consequences > that might not be acceptable for every use case. > Just wanted to make sure you're aware that it's possible to run HiveServer2 in a thick-client mode where it sits embedded in the client JVM process. This allows users to leverage Beeline without having to run/share a standalone HiveServer2 process. Do the performance and scale consequences you mentioned also apply to this mode of operation? - Carl