Under normal circumstances, a 4GB / 2 vCPU server can easily run a 4 unicorns / 1 Sidekiq configuration using only half of the available memory.
That said, if Discourse is going to be an integral part off your application you should be running it in a high available mode, including multiple servers and data centers for resilience against failures. It also a good idea to have it auto scale accordingly to traffic. Those are all characteristics of our hosting at discourse.org.