renamed baseline bench and added separate cluster to cfa benchmark to avoid main thread starvation issue