it makes me wonder, though: if I had a table with 50k rows, updated hundreds of times per second and used in joins throughout the database, is there any way I can just stick that whole table into memcached or redis? I know there are some cases where this works, some where it doesn't. curious if this option was explored.
A relational db is originaly meant for fetching data from a slow storage to a fast one, and the other way around, in a smart way. Doing it 500 times a second isn't a scenario for any db.
Build your own in memory data + process tructure, maybe using something like and agent network and using eg akka or erlang for failovers.
I'm curious as to why uber had to rely on a relational db for this case...
As you say though, put it in memory first and write it out to a DB every now and then.
[1] Source: https://momjian.us/main/writings/pgsql/hw_performance/