I built a plugin for visualizing network topologies at scale.
When observing anything more complex than a small lab setup, most existing tools assume subscription based vendor lock or rigid schemas for setup, hardcoded dataframes, or collapse under visual clutter. I wanted something that works across data stacks, scales with volume, and remains readable.
Key features:
- No hardcoded dataframe requirements. It works with arbitrary data models
- Vendor-agnostic and transferable across datasource stacks
- GPU-accelerated rendering for large graphs
- Layered autolayout with subgraphs (e.g. namespaces) to reduce visual clutter
- Link aggregation for shared path fragments
- Metric reducers (e.g. show only faulty or overloaded connections)
- Dynamic styling with user-defined node groups derived from data
- Cluster icons with aggregated group statistics
- Bi-colored 3D arcs to visualize link load
This is aimed at real-world networks (infra, service meshes, distributed systems), not just small demo graphs.
I’d love feedback from people working on observability, networking, or graph visualization - especially on usability, missing features, or integration pain points.
As for Grafana labs, I've talked to them right at the inception offering cooperation on building a faster geomap with links that only lead to shipping their own Network layer in a rush on top of existing outdated tech stack with no intention to improve, so it never left Beta. Native Grafana Nodegraph is also highly hardcoded for internal needs.
I wonder why people keep bombarding Grafana github issues with feature requests for 3-5 years since major updates of these plugins instead of looking for an alternative?
I also have a community edition, which I stopped maintaining due to lack of volunteer support. It remains listed in the Grafana catalog with over 200k downloads, mostly from users who are fine with a large gap from the current version.