The backend is distributed Actors. Instead of bringing the data to the (web)request, the Actor model brings the request to the data (which may be on another server). Actor objects are loaded in memory and kept there while needed, so its very fast by itself and there's no need for caching at all.

Storage can be very simple, it is possible to use SQL but I'm using Table Storage which is cheaper and more scalable. For searching we use ElasticSearch.

This server model fits very naturally with websockets.