you can often as easily deploy a monolith to multiple servers and grant different subtasks different cpu and memory for their work. for me its just a matter of adjust with the supervisord config for each server, how many of which workers start and run on that machine.