Network Topology Types, Explained for People Who Will Have to Monitor One
Most explanations of network topology were written for an exam, which is why they spend equal time on bus and ring networks that nobody has installed this century. This one is written for the person who has just inherited a network and needs to know what shape it is, what that shape implies when something breaks, and how to draw it so the next person does not have to guess. The exam material is here too; it is just in proportion.
In this guide
- What network topology means
- Physical versus logical topology
- Star topology
- Bus and ring topologies
- Mesh network topology
- Tree and hybrid topologies
- Point-to-point and the WAN
- Which topology is most common, and which is best
- Drawing the topology you have
- What topology means for monitoring
- Where to go next
Explain network topology: the definition (“def”) and the explanation that matters
Network topology is the arrangement of the devices on a network and the links between them: which box connects to which, and through what. The definition that matters operationally is narrower than the textbook one. It is the set of things that fail together. When a core switch dies, every device plugged into it is unreachable, not because they failed but because the topology put them behind it. Every type of network topology below is really a description of what depends on what, and that is exactly the information a monitoring system needs to raise one alert instead of forty.
The difference between physical and logical network topology
The physical topology is the cabling: which port of which switch the server's cable actually runs to. The logical topology is how traffic flows: VLANs, routed subnets and tunnels that may cut across the physical layout entirely. Two servers in the same rack on the same switch can be in different VLANs and logically far apart; two branch offices on different continents can be one logical hop apart over a VPN. Documentation needs both, and most networks only have the second one, drawn from memory, three reorganisations ago.
Star topology: the one you almost certainly run (advantages, pros and cons)
In a star network topology every device connects to a central switch. Modern Ethernet is a star by construction, because a switch port is a point-to-point link, so every office network, every branch, every server room is a star or a tree of stars. The advantages of a star topology are the reason it won: one cable fault takes out one device; adding a device is plugging in a cable; and the central switch can see every conversation, which is what makes SNMP port counters and port mirroring possible. The disadvantage is the obvious one: the centre is a single point of failure, and the whole building goes quiet when it fails. Pros and cons, then: easy to build, easy to monitor, one critical box.
Bus and ring topologies: where they went
A bus topology runs every device off one shared cable, with terminators at the ends; it was 10BASE2 coaxial Ethernet, and one loose connector took the whole segment down. A ring topology passes a token from device to device around a loop; it was Token Ring and FDDI. Both are gone from the LAN. The ring survives in two places worth knowing: carrier and metro fibre, where a ring with protection switching gives two paths to every site, and some industrial Ethernet protocols on the plant floor, where a ring of switches tolerates one cut. If you meet a ring today, it is probably one of those, and the thing to monitor is whether the protection path is still intact.
Mesh network topology: full, partial and wireless, and its advantages
In a mesh network topology devices connect to several others, so there is more than one path between any two points. A full mesh connects everything to everything, which costs n(n-1)/2 links and is only done between a handful of core devices. A partial mesh connects each device to a few neighbours; the public internet is a partial mesh of networks, and so is a well-built data centre spine. Wireless mesh is the consumer meaning: access points that relay to each other without cables, trading bandwidth for convenience. The property that matters is redundancy: a mesh survives a link failure without anyone noticing, which also means failures go unnoticed until the second one. Monitoring a mesh means monitoring the links, not just the devices.
Tree, two-tier and three-tier topologies: core, distribution, access, collapsed core, and the hybrid topology
A tree (or hierarchical) topology is stars stacked on stars: access switches in the wiring closets connect up to distribution switches, which connect up to a core. It is how any network bigger than one switch is built, and the three-tier naming is the vocabulary every vendor uses. A collapsed core merges the core and distribution tiers into one pair of switches, which is the right design for most sites below a few thousand ports. A hybrid topology is whatever you actually have: a tree of stars at headquarters, a partial mesh between the core switches, point-to-point WAN links to the branches, each branch its own small star. Every real network is a hybrid; the name is only useful on the exam.
Point-to-point, and the WAN
A point-to-point topology is one link between two devices: a leased line, a fibre between two buildings, a site-to-site VPN. Wide-area networks are made of them, arranged either as a hub-and-spoke (every branch to headquarters, which is a star again at a larger scale) or as a partial mesh when branches need to talk to each other directly. The thing to document about a WAN link is not its shape but its two ends and who is responsible for the middle, because that is what the ISP will ask when it fails.
The different types of network topology compared: which is most common, most popular, and best
The most common network topology by a wide margin is the star, stacked into a tree, with a mesh only between the core devices and point-to-point links to anywhere remote. The "best" topology is a question that only makes sense against a failure you are trying to survive: a mesh survives a link cut, a star does not; a star is cheap and simple, a mesh is neither. For most sites the practical answer is a tree with a redundant core pair and two uplinks from every access switch, which is a star for the clients and a small mesh for the backbone. The table puts the trade-offs side by side.
| Topology | What fails together | Cost of adding a device | Where you will see it |
|---|---|---|---|
| Star | Everything behind the central switch | One cable | Every LAN segment |
| Tree / hierarchical | Everything below the failed tier | One cable, maybe one uplink | Any multi-switch site |
| Mesh (full or partial) | Nothing, until two things fail | Several links | Core pairs, data-centre spines, the internet |
| Point-to-point | The two ends lose each other | A whole new link | WAN, inter-building fibre, VPN |
| Bus / ring | The whole segment, or one direction of it | A segment outage, historically | Legacy; carrier and industrial rings |
Drawing the computer network topology you have: diagram makers, topology mappers, pictures, images and icons
A network topology diagram drawn by hand records what someone believed on the day they drew it. The useful version is generated from the network itself: switches announce their neighbours over LLDP and CDP, and a tool that listens on the wire can draw the physical topology, port by port, and keep it current as cables move. Hand-drawn diagrams still have a place for the logical layer (VLANs, subnets, routing) and for the WAN, where there is no wire to listen to. The network diagrams guide covers what a drawn map should and should not try to show, and the auto-generated maps guide covers the LLDP/CDP approach. Use consistent icons, label every link with its speed, and date the drawing; an undated diagram is a rumour.
Ethernet network topology, subnetting and the various topologies of a wireless network
Switched Ethernet is a star at every switch and a tree between them, whatever the cabling looks like in the closet; subnetting layers a logical topology on top, one subnet per VLAN or per site, which is what the routers see. A wireless network is a star around each access point (infrastructure mode), a mesh between access points that relay to each other, or point-to-point between two bridges across a car park.
Free and open-source network topology mappers, and what a topology mapper download gives you
A topology mapper listens for LLDP and CDP and reads switch tables over SNMP to draw the physical map, which is a different job from a diagram maker you draw in. The open-source ones do it well for a lab; a monitoring system does it continuously and uses the result for alerting. The auto-generated maps guide covers both, and the discovery and mapping page covers how SysPeep does it for up to 5 devices free.
What topology means for monitoring: dependencies and alert storms
This is the reason an operations person should care about any of the above. A monitoring system that knows the topology knows that the forty devices behind a dead switch are not forty failures; they are one failure with forty consequences. Without that knowledge it pages you forty times and you turn the alerting down, and then mostly off. With it, the switch raises one incident and the forty are listed under it as unreachable. The dependency map is the topology, discovered from LLDP and CDP rather than typed in, and it is the single most valuable thing a monitoring tool can learn about your network on its first afternoon. The alert correlation and alert fatigue series take it from there.
Where to go next
If the network is undocumented, start with the undocumented network series and let discovery draw the first map before anyone draws it by hand. The network documentation pillar covers what to write down and how to keep it true. For the devices that make up the topology, the switch port monitoring and router and WAN monitoring series explain what to watch on each. And the product side of discovering and mapping a network from the wire is on the network discovery and mapping page.
Put this into practice with SysPeep Monitor
SysPeep draws the topology from LLDP and CDP on the wire, keeps discovering as a service after the first sweep, and uses the same dependency map to roll the devices behind a failed switch under its one incident. Up to 5 devices are free at full depth, so a small site can be mapped and monitored for nothing.
Download Free TrialLearn About SysPeep