Preparing a four-node cluster

To get started, we need three or four Linux nodes connected on a common network where we can install Docker Enterprise and host our PoC. Keep in mind that for our PoC, one node is for UCP, another for DTR, and the remaining nodes are workers. A minimum of a four-node setup is highly recommended for PoCs to demonstrate multiple workers, and hence an actual orchestrated cluster. In Kubernetes speak, we would have one master and three nodes, where one node runs DTR and the remaining nodes run containerized workloads. 

Please keep in mind, this is a PoC and does not reflect your final platform for Docker Enterprise; preliminary decisions on your actual target platform are not required until you move to the pilot step. These decisions are intentionally deferred because, when you are getting started with Docker Enterprise and in the PoC step, you lack the real firsthand knowledge and experience required to make informed longer-term decisions. Furthermore, trying to force these decisions during the PoC adds to the learning curve and introduces more rookie mistakes, resulting in costly delays and potentially institutionalizing anti-patterns. So, start in your comfort zone by choosing a Docker Enterprise supported infrastructure that aligns with your current skill sets. Remember, the goal here is to gain some experience with Docker Enterprise as the target platform for your PoC and not to try a bunch of new things at the same time.