Network Topologies

Once you connect two or more computers together you have a network. But there is more than one way to wire the machines up — and the shape of those connections matters enormously. A network topology is simply the layout: the pattern of how devices are joined to one another and to the equipment in between.

Think of it like the roads between towns. You could give every town a single road to one big central junction, or you could criss-cross the country with roads running directly between many towns. Both let you get from A to B — but they behave very differently when a road is blocked, and they cost very different amounts to build. Networks face exactly the same choices, and the two layouts you meet at GCSE are the star and the mesh.

The star topology

In a star topology, every device has its own cable running to a single central device — usually a switch (or, on older kit, a hub). No device talks directly to another; every message goes into the switch and the switch forwards it out to the right destination. It is the layout of almost every home and school network you have ever used.

Because each device sits on its own dedicated cable, the star has some lovely properties:

The catch is that everything depends on the box in the middle — and that is the idea the "Watch out!" box below pins down.

The mesh topology

In a mesh topology there is no single central device. Instead, devices connect directly to many other devices, so between any two machines there are usually several different paths. If one link breaks, the data simply takes another route. This makes a mesh extraordinarily resilient: there is no single point of failure.

There are two flavours. In a full mesh, every device is connected to every other device. In a partial mesh, most devices have several connections but not a link to absolutely everyone — a sensible compromise that keeps a lot of the resilience without paying for every possible cable.

Why not use a full mesh everywhere? Cost and complexity. The number of connections grows very quickly. For n devices, a full mesh needs

\text{connections} = \frac{n(n-1)}{2}

links. Ten devices already need 45 separate connections; a hundred devices would need 4950. That is a mountain of cabling (or wireless links) and hardware, which is why full wired meshes are rare — but wireless mesh networks, where each node just needs a radio, are increasingly common.

Pretty much — and that is no accident. The internet grew out of a military-funded network (ARPANET) whose whole design goal was to keep working even if parts were destroyed. By wiring routers together mesh-style with many redundant paths, a message can flow around any break. It is why you rarely notice when a single cable is cut somewhere on the other side of the planet: the packets simply reroute. The internet is not a full mesh (that would be impossible at global scale) — it is a vast partial mesh of networks.

Star vs mesh: the trade-offs

Neither layout is "best" — each wins on some measures and loses on others. Engineers choose based on how much they care about reliability, cost and performance for a particular network.

MeasureStarMesh
Reliability A single cable/device failing only affects that device — but the whole network dies if the central switch fails (a single point of failure). Very resilient: multiple paths mean no single point of failure. A broken link just reroutes.
Cost Cheap: one cable per device plus one switch. Expensive: many cables/links and more hardware, especially a full mesh (n(n-1)/2 links).
Performance Good — the switch routes each message directly; easy to add devices without slowing others. Good — traffic can spread across many paths, avoiding congestion, though managing all those routes is complex.

Rule of thumb: use a star when you want something cheap and simple (a home, an office, a classroom); reach for a mesh when staying up matters more than saving money (the internet's backbone, critical wireless networks).

Don't muddle up the two different kinds of failure in a star.

A mesh has no such weak spot: with many paths between devices, losing any one link (or even a node) just reroutes the traffic. That extra resilience is exactly what you are paying for with all those extra connections.