It just means that hash collisions will be more frequent above the capacity, causing more walks on hash chains. It is unclear at what level you will see meaningful impact. In any case, it is not really a limit.
Kihwal On Tue, Jun 25, 2019 at 2:17 AM Lars Francke <lars.fran...@gmail.com> wrote: > Hi, > > I stumbled upon the "Block capacity" metric in HDFS and understand that > it's the capacity of the internal block map and that it's somehow tied to > the heap size (2%). > > I've looked at the code for LightWeightGSet (admittedly only briefly) and > it says that it's using an array + linked lists for conflict resolution. > > So, having seen that are we really limited by the "capacity" of the block > map? > > Cheers, > Lars >