If you are running your backend on cloud, then as a rule of thumb, you never use clustering. You deployment takes care of it. AWS, K8S autoscaling takes caring of spinning up new instances of your application and allocating it VM resources. And for worker_threads you use it for tasks that are going to stay in the memory for very long. For example if you have a basic CRUD server, no need to use worker_threads but if you have to make a very lofty query and get a lot of data from the DB and then process that data, make changes, write to a file, upload that file, send a response back, you use worker_threads.