Scales to very high numbers of connected clients, and is faster on the low end and faster on the high end. Beating both the default per-thread and the Enterprise Edition thread_pool connection handler. Enjoy!