Stacks, services, and tasks

When using a Docker Swarm versus a single Docker host, there is a paradigm change. Instead of talking of individual containers that run processes, we are abstracting away to services that represent a set of replicas of each process, and like through become highly available. We also do not speak anymore of individual Docker hosts with well known names and IP addresses to which we deploy containers; we'll now be referring to clusters of hosts to which we deploy services. We don't care about an individual host or node anymore. We don't give it a meaningful name; each node rather becomes a number to us. We also don't care about individual containers and where they are deployed anymore—we just care about having a desired state defined through a service. We can try to depict that as shown in the following figure:

Containers are deployed to well known servers

Instead of deploying individual containers to well known servers like the preceding one, where we deploy container...