The original topology view crammed L2 physical links, L3 subnet structure, and service dependencies into one diagram. That worked for small networks, but stopped being readable as they grew. This post covers how we approached the redesign of our visualization engine: we created four separate views (Physical L2, Logical L3, Workloads, Applications), implemented zoom levels inspired by the C4 model for software architecture visualization, and used ELK layered layout (elkjs) for the hierarchical views. Design principles were also inspired by several graph readability papers I cite in the post.
There's a free self-hosted community edition (AGPL) if you want to try it!