There has a large throughout between worker node and master node. Because our node deployed in different network, their communications through public network, it's so expensive. Can we do some optimization for this?