Network topology is the arrangement of nodes and the links between them in a computer network, together with the pattern by which data travels across those links. In networking fundamentals it is the first design decision that shapes cost, resilience, performance and ease of troubleshooting, because it determines how many paths exist between devices and what happens when one link or node fails.
Topology has two views that often differ. The physical topology describes where cables, fibers and radio links actually run; the logical topology describes how frames and packets move between nodes. A modern office network can be a physical star of cables into a switch while behaving logically as a single broadcast domain, or as several VLANs.
The basic topology types
Network engineers describe most designs with a small set of classic shapes. Each has a characteristic number of links, a typical failure behavior and a natural place in history or in modern practice.
| Topology | Structure | Links for n nodes | Effect of one link failure |
|---|---|---|---|
| Bus | All nodes share one cable | One shared segment | A cable break can split or stop the whole network |
| Ring | Each node links to two neighbors in a loop | n | Single ring breaks; dual rings survive |
| Star | Every node links to a central device | n – 1 | Only the attached node is cut off |
| Full mesh | Every node links to every other | n(n – 1)/2 | Traffic takes another direct or indirect path |
| Partial mesh | Selected nodes have multiple links | Between n – 1 and full mesh | Depends on remaining redundant paths |
| Tree | Stars connected hierarchically | n – 1 | Everything below the failed link is isolated |
The link count formula explains why full mesh is rare: ten nodes need 45 links, and fifty nodes need 1,225. Designers therefore use a partial mesh where resilience matters and a star or tree where cost and simplicity matter more.
Physical and logical topology
The physical and logical views answer different engineering questions. The physical view tells a technician which port and cable to inspect; the logical view tells an engineer where broadcasts reach, which path a packet follows and where a failure will actually be felt.
| Aspect | Physical topology | Logical topology |
|---|---|---|
| Describes | Cables, ports, devices and locations | Paths, segments and data flow |
| Typical documentation | Rack diagrams, floor plans, patch records | VLAN maps, routing diagrams, IP plans |
| Changed by | Recabling or new hardware | Configuration, VLANs, routing policy |
| Example | Star of copper cables into one switch | Two VLANs sharing that switch |
Classic Ethernet over coaxial cable was a physical and logical bus. Today Ethernet runs as a physical star, yet a group of switches still forms a logical Layer 2 domain where loops must be prevented by the Spanning Tree Protocol or by routed designs. Wireless LANs add a further layer, because radio links create a logical star around each access point with no visible cabling at all.
Topologies in modern networks
Real networks combine basic shapes into hybrid topologies suited to their scale. The patterns below are the ones most often found in current LANs, data centers and wide area networks.
- Hierarchical campus design: access, distribution and core layers form a tree with redundant uplinks, which is a partial mesh at the upper layers.
- Spine-leaf data center fabric: every leaf switch links to every spine switch, giving equal path length and predictable latency between any two servers.
- Hub-and-spoke WAN: branch offices connect to a central site; it is cheap and simple, but branch-to-branch traffic must cross the hub.
- Meshed WAN or SD-WAN overlay: branches build direct tunnels where traffic justifies them.
- Wireless mesh: access points relay traffic for each other where cabling is impractical.
In the spine-leaf model, a server reaches any other server in at most three switch hops, and links are used in parallel through equal-cost multipath routing. That property explains why it replaced the older three-tier tree in most large data centers, where east-west traffic between servers dominates.
Choosing a topology
Selecting a topology is a trade-off between redundancy and cost, moderated by the traffic pattern and the skills of the people who will run it. The decision table below summarizes common situations.
| Situation | Recommended topology | Reason |
|---|---|---|
| Small office of 10-50 devices | Single star around one switch | Low cost and simple management |
| Multi-floor campus | Hierarchical tree with redundant uplinks | Scales cleanly and contains failures |
| Data center with heavy server-to-server traffic | Spine-leaf fabric | Uniform latency and parallel paths |
| Retail chain with many branches | Hub-and-spoke or SD-WAN overlay | Central control of security and policy |
| Critical core between a few sites | Full or partial mesh | Survives multiple link failures |
A useful test is to ask what single failure can stop the business. Any device or link whose loss disconnects many users is a single point of failure, and the topology should either remove it with redundancy or accept it consciously because the cost of removal is too high.
Common mistakes in topology design
Many network outages trace back to topology decisions, not to faulty hardware. The most frequent errors repeat across organizations of every size.
- Hidden single points of failure, such as two redundant links that share one conduit or one power feed.
- Unplanned Layer 2 loops created by adding a cable between switches without loop protection.
- Undocumented growth, where daisy-chained switches turn a star into a fragile, deep chain.
- Oversubscribed uplinks in a tree, where many access ports share one slow link to the core.
- Diagrams that are out of date, which slow every troubleshooting session.
Keeping both physical and logical diagrams current is the cheapest protection against these problems. Automated discovery through LLDP and network management tools helps, but it does not replace a deliberate design.
A non-technical view
Network topology is the map of how devices are connected. Think of roads between towns. If every town connects only to one central city, the map is simple, but a closed road around the city cuts many towns off. If towns are connected to each other directly, there are always alternative routes, but building all those roads costs much more.
Computer networks face the same choice. Designers pick a layout that gives enough backup paths for the things that must not stop, while keeping the rest simple and affordable.
A concrete example
A hospital with three buildings runs its clinical systems over a campus network. The original design was a simple tree: each building’s switches connected over one fiber link to a core switch in the main building. When an excavator cut the fiber to the outpatient building, the whole building lost access to patient records for six hours.
The redesign keeps a tree inside each building but adds a second core switch in another building and a second fiber route along a different path. The two core switches link to each other, and every building has one uplink to each core, forming a partial mesh at the core. In a later test, engineers disconnected one core switch during a maintenance window, and traffic shifted to the other path within seconds, with no interruption to clinical applications.