I mentioned that I did database design. From early readings of Codd and Date nearly two decades ago, have spent a professional lifetime grappling with the practical implementations of third normal form relational models and also their "modern" multidimensional variants. The process of mapping real world phenomena to basic binary patterns is to me the single most important factor in the success of any computer system, so it's no surprise that I genuinely enjoy it.
As with the central nervous system of any organism, as with the human brain, as with consciousness itself, it is the pattern of connections that defines its being and purpose - it is the topology of the neural network that determines who we are, not the physical implentation of that pattern. You could lose any nerve in your body, but if that nerve were replaced by an equivalent electrical transmitter with the same trigger and firing mechanism, then your essence would be unchanged, your personality would be unchanged, your thoughts would be unchanged. Every thought that we have is the process of creating new connections.
But as it is topology that counts, on a data model or a tube map or a nervous system, the representation of that network on a two dimensional sheet can be shown in any number of apparently different ways, but so long as the relationship of nodes to connectors is the same then the processing will be the same. It does not matter to the logic of an electrical circuit if the wires are layed out differently. However, where art meets science, it matters to me. Certain patterns of symmetry and chaos are considered more pleasing than others, possibly where those resonate with basic natural patterns. So the "look" of a data model matters - not so much as its topology, but it still matters.
So network design is a massive subject, one that can't really be summarised within brief guidelines. But top five principles of representing network design to follow shortly.