NoSQL cluster worked fine except for two minor things: reading and writing data

This is a great article. It really shows the negative effects of going into a horizontally scaled db system for the performance gains. One part I will note is that, if you dont care about nanosecond consistency, is that mongo does allow you to scale reads by reading from a secondary versus the primary. Frees up the primary to perform all writes. Writing is still a bottle neck.

But as the article states, if you keep the entire db in memory, why not. Way faster. Just make sure it rights to disk fast so no data is lost in a hardware or power failure.