Reading networks from the center
An independent reference on network structure: what hubs are, how hub-and-spoke systems are laid out, and how centrality is measured.
Not a company site, not a product page, and not affiliated with any business of a similar name. Nothing here is sold, booked, or offered.
A hub is a node that other nodes depend on — because they attach to it, pass through it, or reach the rest of the network by way of it.
Every network is points and links. Most points are interchangeable: remove one and the network barely notices. A few are load-bearing — remove one and trips lengthen, flows reroute, whole regions lose contact with each other. The load-bearing points are hubs.
The obvious test, counting links, is a real measure but an incomplete one. A node can be the most connected point in a dead end, while a modest three-link node holds two halves of the map together. The first is popular; the second is central. This dossier keeps the two ideas separate on purpose.
Three signatures recur wherever hubs appear. The hub collects: flows from many points converge on it. It redistributes: arrivals are sorted and sent onward. And it abbreviates: any two of its neighbors are two steps apart through it, however far apart they lie on the map.
One center, many rays: everything moves to the middle, gets sorted, and moves out again.
In transport the center is the hub airport: flights arrive in timed banks, passengers swap planes, the bank departs. In logistics it is the sorting hall, where every parcel travels whatever its origin. In computing it is the star topology: one cable per device, and a switch that forwards each frame to its destination.
The arithmetic is the layout's main argument. Connecting n points directly to one another takes n(n − 1)/2 links; connecting them through a center takes n − 1. Ten points mean 45 direct links against 9 through a hub — and the gap widens with every point added.
Read from the other side, the same arithmetic is the weakness. A pure star has exactly one path between any two spokes, and it runs through the middle. Real systems therefore cheat on the purity: a second hub, a fast spine between hubs, direct routes between the busiest spokes.
Degree counts a node's links, closeness measures its average distance to everyone else, and betweenness counts how often it sits on the shortest path between others.
Degree centrality is the simplest count: how many links touch this node. It answers a local question — how much of the network is within one step — and it cannot see the wider map, so it regularly over-ranks busy nodes whose many neighbors lead nowhere.
Closeness centrality asks a global question: from this node, how many steps does it take, on average, to reach every other node. High closeness marks a good broadcaster, because nothing is far away from it. It rewards being in the middle of the web rather than on top of a pile of links.
Betweenness centrality asks a different global question: how often does this node lie on the shortest route between two other nodes. High betweenness marks a broker or a bottleneck — a node that may have few links yet carries the traffic of two halves. On one and the same graph, the three measures regularly name three different nodes as the most central.
The list points to the classic papers and standard textbooks behind the claims on this page; it is a reading list, not an endorsement.