Skip to content

Overview

The overview is where a session starts. It answers one question at a glance: what is my network talking to right now, and is anything wrong?

A dark globe with your site as a single point and every active external connection as an arc from it, drawn in the direction of the data.

Element Encodes
Arc thickness Bytes
Arc colour Live traffic — one colour, not a severity scale
Arc opacity Raised by a high or critical incident on that path
Home point Your configured site coordinates
Ring on an arc A destination never seen before, held for ten seconds

The colour is worth a sentence, because it is a deliberate constraint rather than an oversight. Orange on the globe means “traffic is flowing”, not “this is bad”. A serious incident raises the opacity of an arc, never its hue, so the picture never turns into a threat-level map that is impossible to read at a glance.

Beside the globe sits a list of countries and networks with device counts and byte totals. Clicking an entry zooms the globe to it.

The globe rotates slowly and stops on interaction. A quiet switch turns off rotation, particles and pulses entirely — for a wall display, or for a laptop without a GPU.

Three ordinary reasons, in order of likelihood:

  1. No GeoIP databases. Without them the geo fields are null and the globe draws nothing, because it does not invent locations. See GeoIP.
  2. No home coordinates. A tenant without configured coordinates has no origin to draw from.
  3. No external traffic in the window. The default window is the last 15 minutes.

Down one side: the open incidents, sorted by score. Each card carries its title, severity, the affected devices, and a short summary. Clicking one opens the incident dialog without leaving the view, and the URL changes with it, so an incident is a link you can send to someone.

On a fresh installation this rail is empty on purpose. See the learning phase.

Always visible: how fresh the data is, and whether every source is delivering. It links into the system view.

This is not decoration either. A dashboard that looks the same whether or not it is receiving data is a dashboard that lies. If a source is stale, the overview says so before you draw any conclusion from what it is showing.

The view is bounded so that it stays legible and cheap:

  • at most 300 arcs on the globe, with the remainder aggregated per network,
  • at most 500 nodes and 2,000 links in the network map,
  • a total particle budget, above which small flows are drawn without particles.

When aggregation happens it is explicit in the data — a cluster node says how many members it stands for — rather than silently dropping the rest.

Updates arrive as deltas over a server-sent event stream: nodes and links are mutated rather than rebuilt, so the layout stays still while the data moves.