This presentation from 2019 gives an idea about some of the GC avg/max runtimes using the NameNode: https://www.slideshare.net/xkrogen/hadoop-meetup-jan-2019-dy...