How it works
Each instance gets its own container that contains its database services, the program language interpreter, and its own web server.
Client requests arrive at a load balancer that distributes traffic to http proxies. These proxies benefit from a very efficient local cache: for each request, they verify whether or not there is a more recent version already in the cache. If it is present, the reply is automatically sent to the client. The cache’s “time to live” depends on the “max-age” http header.