Expected Behavior

Complex topologies with interwoven links and 100s of devices are difficult to parse and organize when viewing all nodes especially in mixed network environments with fat tree, S&L, rails, and other fun paradigms in one view. Some graph viewers offer the ability to force-direct node positions, define repulsion by node edge/attribute scaling (such as links to various devices), and apply various algorithms for placement to generate topological views not requiring manual intervention for positioning which would be of help in these device forests.

Current Behavior

Currently the neighbor topology view does its best to represent what it sees but gets extremely cluttered and convoluted in complex environments while also (infuriatingly) forgetting painstakingly placed node locations after some time forcing engineers to spend hours pinning cores, spines, leaves, rails, routers, peering points, and other logic to XY coordinates on the graph.

Possible Solution

Aside from providing options for applying algorithmic organization/pre-formatting of the node graph, the ability to save and restore positions would be incredibly useful even as new nodes are added or existing ones change. Delta visualization from a saved state would also be fairly handy and any mechanisms for topological classification and ordering (spines, leaves, rails, fat trees, or dragonflies if IB becomes a thing NetDisco can handle) based on link patterns or identified patterns of configuration attributes would be phenomenal.