What is a network topology?
A topology is the shape of a network: which devices connect to which, and how data travels between them. It is like a map of the cables (or radio links) drawn without worrying about exact distances or which room things are in.
Each device on the map is called a node, and each connection is a link.
Why topology matters
The shape of a network decides a lot about how it behaves:
- Reliability: what happens when one cable or device fails? Does one device go offline, or everyone?
- Cost: how many cables, ports and devices do you need to buy?
- Performance: does traffic share one cable, or does each device get its own?
- Growth: how easy is it to add the 51st device?
- Troubleshooting: how quickly can you find the broken part?
💡 In simple terms: topology is like the road layout of a town. One main street (a bus), a roundabout with roads going out (a star), a ring road, or a grid where every junction connects to many others (a mesh). Each layout handles traffic jams and road works differently.
Physical vs. logical topology
Every network has two topologies, and they don't have to match:
| Physical topology | Logical topology | |
|---|---|---|
| Describes | How the cables and devices are actually connected | How the data flows between devices |
| You find it by | Following the cables, looking at the rack | Understanding how the devices forward data |
| Example | Every PC is cabled to one box in the cupboard: a star | If that box is a hub, every signal goes to every PC: it behaves like a bus |
Classic examples: a hub network is a physical star but a logical bus; an old Token Ring network was cabled as a star but passed data around a logical ring; and Wi-Fi is a physical star around the access point where all devices share the same radio channel.
Point-to-point
The simplest topology: one link joining exactly two devices. Whatever one end sends, only the other end receives. Nothing to share, nothing to decide.
- 1. One link, two ends. Everything sent by the head-office router arrives at the branch router, and nowhere else.
- 2. Both directions. The branch replies over the same dedicated link.
Every cable from a PC to a switch port is itself a point-to-point link. That's why modern switched Ethernet has no collisions: each link has only two ends. See Collision domains.
Star
Every device connects with its own cable to one central device, today usually a switch (or a Wi-Fi access point). Devices never connect to each other directly; everything goes through the centre.
- 1. Through the centre. PC A sends to the server. The switch forwards it only out of the server's port.
- 2. At the same time. Laptop B prints at the same moment, on different ports, without getting in A's way.
- 3. One cable fails. PC A's cable is cut: only PC A goes offline. Everyone else carries on.
- 4. The centre fails. If the switch itself dies, every device loses the network. The centre is a single point of failure.
Bus
All devices attach to one shared cable, the backbone or bus. A signal sent by one device travels the whole length of the cable, so every device hears every transmission. Only one device can send at a time.
- 1. Send. PC B puts a frame for PC D onto the shared cable.
- 2. Spread. The signal travels both ways along the whole backbone.
- 3. Everyone hears it. Every PC receives the signal. Only PC D, the destination, keeps the frame; the others ignore it.
- 4. Absorb. A terminator at each end soaks up the signal so it does not bounce back and garble the next one.
- 5. Collision. If PC A and PC C send at the same moment, their signals crash into each other on the cable and both must try again later.
Ring
Each device connects to exactly two neighbours, forming a closed loop. Data travels around the ring from device to device until it reaches the destination. Many rings send in one direction; dual rings have a second ring going the other way as a backup.
- 1. Around the ring. Site A sends to Site C. The data passes through Site B on the way.
- 2. Cable cut, go the other way. The A–B link is cut. Because it's a ring, traffic can still reach C the other way round, through D.
- 3. Two cuts split the ring. With a second cut (C–D), Sites B and C are isolated from A and D. A ring survives one break, not two.
Mesh
In a mesh, devices have several links to other devices, so there is more than one path between them. If one path breaks, traffic takes another. There are two kinds.
Full mesh
Every device has a direct link to every other device.
- 1. Direct path. Every router has a direct link to every other one: R1 sends straight to R4.
- 2. Link fails, reroute. The R1–R4 link fails. Traffic goes via R2 (or R3) instead. Users barely notice.
- 3. Still another way. Even with R2 also out of action, R3 offers a third path.
The catch is the number of links. For n devices a full mesh needs n × (n − 1) ÷ 2 links, and every device needs n − 1 ports:
| Devices | 4 | 5 | 10 | 20 | 50 |
|---|---|---|---|---|---|
| Links needed | 6 | 10 | 45 | 190 | 1,225 |
Partial mesh
Only some devices have extra links, usually the most important ones. It gives most of the reliability for a fraction of the cost, and it is how real networks are built.
- 1. Important site, two links. Branch 2 is large, so it has links to both HQ and the data centre.
- 2. Backup path. Its direct link fails, so it reaches the data centre through HQ.
- 3. Small sites, one link. Branches 1 and 3 have a single link each. A failure there cuts them off, which the business has decided is acceptable.
Spare links create loops. Routers handle loops naturally, but loops between switches need a protocol called Spanning Tree to stay safe. That is a CCNA-level topic: see Spanning Tree if you want to go further.
Hybrid
A hybrid topology combines two or more of the shapes above. Almost every real network beyond a single room is a hybrid. The most common pattern is a star of stars (also called a tree or hierarchical design): each floor is a star around its own switch, those switches connect in a star to a core switch, and the core links to other sites with a partial mesh.
- 1. Star on each floor. Floor 1's devices connect in a star to their floor switch.
- 2. Star of stars. The floor switches connect in a star to the core, so any floor can reach any other.
- 3. Meshed edge. Two WAN routers, linked to each other and the WAN, give a partial mesh at the most important point.
- 4. Survives a router failure. If WAN router 1 fails, traffic simply leaves through router 2.
All topologies compared
| Topology | Cabling cost | One link fails | Single point of failure? | Where you see it today |
|---|---|---|---|---|
| Point-to-point | Lowest | The two ends are cut off | The link | WAN links, uplinks |
| Star | Medium | One device offline | Yes: the central device | Home and office LANs, Wi-Fi |
| Bus | Low | Whole network down | Yes: the backbone | Legacy; vehicle and industrial buses |
| Ring | Medium | Dual ring: reroutes. Single ring: down | Two breaks split it | City fibre rings, provider networks |
| Full mesh | Very high | Reroutes | No | Small groups of core routers, data centres |
| Partial mesh | High | Usually reroutes | Only for single-linked sites | WANs, the internet core |
| Hybrid | Varies | Depends on the part | Designed out where it matters | Almost every real network |
A real-world example: a school
A secondary school has three buildings. Each classroom has PCs and a ceiling Wi-Fi access point cabled to a switch in a cupboard (a star per floor). Each cupboard switch has a fibre uplink to the main switch in the server room (a star of stars). The server room has two internet connections from two different providers, each on its own router, linked to each other (a partial mesh at the edge). The whole design is a hybrid. When a classroom cable is damaged, one PC goes offline; when one provider fails, the school keeps working through the other.
When topology causes problems
Most outages trace back to a single point of failure: one device or link that everything depends on. Topology tells you where they are:
- Star: a whole floor goes down → check the floor switch, its power and its uplink before checking PCs.
- One user down in a star: check that user's cable, wall socket and switch port light.
- Ring or mesh: users notice a short pause, then everything works but slower → a link failed and traffic took a longer path.
- Loops by accident: someone plugs both ends of a cable into the same switches → the network can flood itself. This is why switch loops need care.
You can watch a mesh reroute from a PC. Here a continuous ping (-t) runs to a server at another site while a WAN link fails:
C:\>ping -t 10.2.0.10 Pinging 10.2.0.10 with 32 bytes of data: Reply from 10.2.0.10: bytes=32 time=12ms TTL=62 Reply from 10.2.0.10: bytes=32 time=12ms TTL=62 Request timed out. Request timed out. Reply from 10.2.0.10: bytes=32 time=19ms TTL=61 Reply from 10.2.0.10: bytes=32 time=18ms TTL=61 Ping statistics for 10.2.0.10: Packets: Sent = 6, Received = 4, Lost = 2 (33% loss), Approximate round trip times in milli-seconds: Minimum = 12ms, Maximum = 19ms, Average = 15ms Control-C ^C
tracert 10.2.0.10 to see the new path; more in Ping and Traceroute.💡 The best troubleshooting tool for topology is an up-to-date network diagram. Without one, you are guessing which switch a user hangs off.
Common mistakes
- Assuming the picture matches the data flow. A hub network looks like a star but behaves like a bus. Always ask about the logical topology too.
- Thinking a star has no single point of failure. It does: the central device.
- Adding redundant links between switches without planning. Extra links make loops; switches need loop protection to use them safely.
- Calling every multi-path network a full mesh. Nearly all real ones are partial meshes.
- Treating Wi-Fi as point-to-point. Wireless devices share the same radio channel around the access point, a bit like a bus.
- Topology is the shape of a network: physical (cables) and logical (data flow) can differ.
- Point-to-point joins two devices; star joins many through one central device.
- A bus shares one cable (legacy); a ring passes data neighbour to neighbour.
- A mesh gives several paths: full mesh links everything, partial mesh only what matters.
- A full mesh needs n(n − 1)/2 links, which is why most networks use a partial mesh.
- Real networks are hybrids: stars on the edge, meshes at the core and WAN.
Knowledge check
In an office star network, the switch loses power. What happens?
How many links does a full mesh of 6 routers need?
PCs are each cabled to a central hub. Every frame is repeated out of every port. What are the physical and logical topologies?
A company wants its two biggest sites to survive a WAN link failure, but can't afford spare links to all 40 small branches. Which design fits?
Where to go next
You've now seen what networks are, their sizes and their shapes. The next unit looks at the boxes that build them, starting with Network interface cards (NICs), Hubs and Switches. For how large LANs are laid out in layers, see Types of LANs, and for the shapes inside a data centre see Data centre networking basics.