I am considering using Consul to perform server discovery in a WAN federation of many small embedded devices (Raspberry Pi 3 with 1GB RAM and 2GB disk space) connected to a common cloud.
Each device should be able to perform remote service discovery and local service discovery even in case of an interruption of internet service, always giving priority to local services. Therefore I have decided to model each device as a single datacenter.
My questions are:
Is this feasible? In such a constrained environment I am worried that once the federation reaches a certain size, even if all devices are isolated from eachother using Consul Enterprise, the limited amount of resources available will not be enough.
Does this make sense? I am considering this architecture because I can not see another way of implementing these requirements, since I am very new to this maybe there is a much better way of implementing it that I am not seeing.
How much resources do consul servers and cliens use? I have been seeing numbers of between 90 and 200 MB, but in the documentation a machine of 16GB of RAM is recommended for a small server, where does this difference come from?
Thank you very much for your help,