Fri Feb 21 07:11:34 EST 2020

distributed systems and local memory

Conclusion for now is that it's hard to design distributed systems
without some form of local store, and really, you don't want to

In general, communication performance cannot be ignored for
distributed systems, so it is probably best to design it around tasks
knowing the physical locations of the data they operate on.  This
breaks abstraction somewhat, but not doing this makes code very hard
to debug.