I like dtplyr. It shares the vast majority of code with dplyr and runs fast.