A ring topology connects every node to exactly two neighbors, so the cabling forms a closed loop. Data travels around it hop by hop, and every node regenerates the signal. It survives outdoors because it gives two paths to every node for one cable run per node.
- Every node connects to exactly two neighbors
- Every node is also a repeater, so distance behaves differently
- Token passing made it collision free, and switched Ethernet made that moot
- A protected ring recovers in under 50 ms, spanning tree takes seconds
- Alive in metro Ethernet, industrial networks and campus fiber
On this page
How it worksHow data flows around the ring
Each node has two connections, one to the neighbor on each side, and there is no central device. That absence is what separates this topology from the others: a star topology has a hub or a switch in the middle that every conversation passes through, and a ring network has nothing in the middle at all.
In a unidirectional ring, data travels one way around the loop, hop by hop. A node receiving a frame that is not addressed to it regenerates the signal and passes it to the next node, which is why distance behaves differently here than on a shared cable.
Every hop is a fresh signal, so a ring network can span far more physical distance than a single unrepeated run allows.
That same property is the cost. Data between two nodes crosses every node between them, so latency grows with the number of hops and every node in the path is one that can drop what it was given.
The delay is at least predictable: in this topology the worst-case latency can be calculated from the node count rather than estimated, which is one reason industrial networks like it.
A bidirectional ring changes that arithmetic. Data can flow either way around the loop, so the network sends each frame the short way and the worst case halves. Bidirectional operation is also what makes the resilience work, because the second direction is the second path.
Token passingToken passing, and the performance argument it won
The classic access method in this topology is token passing, and it solved a real performance problem that no longer exists in the same form.
A small control frame called the token circulates continuously around the topology. A node may transmit only while holding the token, so two nodes can never transmit at once and there are no collisions at all.
When the sending node has finished, it releases the token to the next node in the ring, and every node on the network gets a turn in a fixed order.
The result is deterministic performance. A shared Ethernet network of the same era degraded unpredictably as traffic rose, because collisions increased with the data offered and the recovery from them made the problem worse.
A token ring at the same utilization stayed orderly, and every connected device was guaranteed its turn within a bounded time. That predictability was the topology's main advantage and it was a real one.
One detail explains most of the confusion about this topology. Token Ring devices did not cable to each other in a physical loop. They plugged into a central hub, the multistation access unit, which wired the ring inside itself.
The topology was a logical ring and a physical star, which is why a Token Ring floor plan and an Ethernet floor plan looked identical.
Two things ended it. Switched Ethernet removed collisions altogether by giving each device its own segment, which delivered the same determinism without the token machinery. And the price of Ethernet hardware fell far enough that being technically better stopped being an argument anybody won.
Where it livesWhere this topology is still the right answer
This is what the textbook advantages and disadvantages lists leave out. A ring topology is not a museum piece; it is what most fiber outside a building actually looks like.
Metro and carrier Ethernet. A provider running fiber past a row of buildings runs it in a loop and comes back. Every site then has two paths to the provider network on one cable run, and a digger through one side leaves every building connected.
Ethernet Ring Protection Switching, standardized as G.8032, is what manages the topology, and it is built to restore the data path in under fifty milliseconds.
Industrial and process control networks. Factory floors run this topology because the cable path is a physical loop around the plant anyway, and because a failure must not stop the line.
Protocols such as MRP and its vendor equivalents recover the network in tens of milliseconds, which is inside what a motion controller tolerates, and that performance is unreachable with general purpose alternatives.
Campus fiber between buildings. Running fiber to eight buildings in a star topology means eight runs back to the core, and in a ring topology it means eight runs between neighbors. The trench is the expense, not the fiber, so the ring network costs roughly half as much and still survives one cut.
SONET and SDH, historically. The telephone network was built on bidirectional dual rings with fifty millisecond protection switching, and that fifty millisecond figure is where the modern Ethernet target came from.
The pattern is consistent. This topology wins wherever the cost of the cable path dominates and resilience is required, which describes almost every network outdoors and almost none inside an office.
The breakThe single point of failure, and the two ways around it
One break in a unidirectional ring splits the loop into a line, and depending on the protocol either half the nodes lose reachability or the whole network stops. That fragility is the most quoted disadvantage of this topology and it is entirely real, which is why no serious deployment uses a single unprotected ring.
A bidirectional dual ring. Two rings carrying data in opposite directions. A break in one is covered by the other, and a break in both at the same point is healed by wrapping the data back at the two nodes either side of the break. This is how SONET worked and how most industrial ring networks still work.
A single ring with one link logically blocked. Cheaper and now more common on Ethernet. The physical cabling is a loop, and the protocol keeps one link blocked so the active topology is a line with no loop in it. When any link fails, the blocked one is unblocked and the data flow continues. G.8032 does this deliberately and quickly.
That second approach explains something that confuses people about Ethernet ring networks. Cabling a loop of switches without a ring protocol produces a broadcast storm within seconds, because Ethernet has no hop count to stop data circulating.
Spanning tree will prevent the storm by blocking a link, and it will take seconds to recover rather than milliseconds. A purpose-built ring protocol does the same job fast enough for the deployments that chose a ring in the first place.
Why the LAN leftThe disadvantages that emptied the office LAN
Three practical disadvantages beat every technical advantage the topology had, and all three are about the building rather than the protocol.
Moves, adds and changes. Inserting a device into a ring topology means breaking the loop, which historically meant interrupting every other device. Adding a device to a star means plugging a cable into a free port on the hub or switch. In an office where people move desks constantly, that difference alone settles it.
Fault isolation. In a star topology, a bad cable affects one connected device and the switch port names it. In a ring topology, a bad segment affects the whole network, and finding it means working around the loop. The star fails smaller and diagnoses faster.
Structured cabling. Buildings are wired as home runs to a wiring closet because standards, contractors and floor plans are all built around that model. That physical reality is a star, and the network topology follows the cable rather than the other way around.
None of those arguments applies to a fiber path between buildings or a cable tray around a factory floor, which is exactly why the ring survived there and nowhere else.
PitfallsWhere people go wrong
Thinking ring topology means Token Ring. Token Ring was one implementation of the topology and it is gone. The ring shape is thriving in metro, industrial and campus networks, running Ethernet, and Token Ring was physically a star around a hub in any case.
Cabling a loop of switches without a ring protocol. The result is a broadcast storm, not a resilient network. Something has to block a link, whether that is spanning tree slowly or a ring protocol quickly.
Buying a single ring and calling it redundant. One ring with no protection protocol is less resilient than a star topology, not more. The advantage comes from the second data path being usable, and making it usable takes a protocol.
Ignoring the node count in the latency budget. Every hop adds delay and every node is in the data path. On a long ring this is calculable, and calculating it is better than discovering it.
Assuming spanning tree is fast enough. Rapid spanning tree recovers the network in seconds. A ring protection protocol recovers in tens of milliseconds. For a factory line or for voice traffic that performance difference is the whole decision.
Comparing topologies without naming the constraint. Ring against star is not a general question about network topologies. It is a question about whether the expensive thing is the cable path or the ongoing changes, and the answer flips depending on which.
ComparisonFour topologies, and the row that keeps the ring alive
| Criterion | Ring | Star | Bus | Full mesh |
|---|---|---|---|---|
| Cable runs for n devices | n | n | Shared | n(n-1)/2 |
| Single point of failure | One break splits it | The central switch | Any break kills it | None |
| Adding a device | Breaks the loop | One free hub port | Interrupts the network | Links to everyone |
| Fault isolation | Around the loop | Per port | Very poor | Per link |
| Latency | Grows with node count | One hop | Shared and variable | One hop |
| Cost of resilience | A protocol, no extra cable | A second switch | Not available | Already there |
| Used today | Metro, industrial, campus | Every office network | Nowhere | Data center fabrics |
The row that decides it is the cost of resilience. This topology buys a second data path to every device without a second cable run, and that is the entire reason it is still built where trenches are expensive and downtime is not tolerable.
FAQFrequently asked questions
What is a ring topology?
A network topology in which every device connects to exactly two neighbors so the cabling forms a closed loop, and data travels around it hop by hop until it reaches its destination.
How does data flow in a ring network?
From device to device, unidirectionally in the simple case. Each device regenerates the signal and passes it along, so every connected device on the path is also a repeater.
What is the difference between a unidirectional and a bidirectional ring?
A unidirectional ring carries data one way around the loop. A bidirectional ring can send data either way, which halves the worst-case hop count and provides the second path that makes the topology resilient.
What is token passing?
An access method where a small control frame circulates and only the device holding it may transmit. It makes collisions impossible and gives every device on the network a guaranteed turn.
Is Token Ring still used?
No. The Token Ring product is long gone, replaced by switched Ethernet. The ring shape it used is very much still deployed, running Ethernet.
What are the advantages of a ring topology?
Predictable performance you can calculate rather than estimate, no contention for the network when token passing is used, resilience from a second data path without a second cable run, and a cable path that matches how fiber is actually laid outdoors.
What are the disadvantages of a ring topology?
A single break splits an unprotected ring, performance falls as every added device lengthens the data path, faults are harder to isolate than in a star topology, and inserting a device means interrupting the loop.
What happens when a ring breaks?
On an unprotected ring, the topology splits and some or all devices lose reachability. On a protected ring, the blocked link is unblocked or the data wraps at the break, typically in under fifty milliseconds.
What is a dual ring?
Two rings carrying data in opposite directions, a bidirectional design, so a failure in one is covered by the other. This is how SONET was built and how many industrial ring networks still work.
Do ring networks need spanning tree?
An Ethernet ring needs something to prevent a loop, and spanning tree will do it. A dedicated ring protocol such as G.8032 does the same job in tens of milliseconds instead of seconds.
Where are ring topologies used today?
Metro and carrier Ethernet, industrial and process control networks, campus fiber between buildings, and historically the entire SONET and SDH telephone transport network.
Why did star topology replace ring topology on the LAN?
Because adding a device is one hub port rather than a break in the loop, a fault affects one user rather than the whole network, and buildings are cabled as home runs to a closet anyway.
Is a ring topology faster than a star?
No. A star is one hop through a switch, and a ring is as many hops as there are nodes in between, so raw performance favors the star. The advantage of the ring is resilience per cable run, never speed.
How many cable runs does a ring topology need?
One per node, the same count as the nodes themselves, because each run joins one neighbor to the next and the last one closes the loop.
Keep readingRelated concepts
Read next · Switching Spanning Tree Protocol A loop of switches without a ring protocol is a broadcast storm, and this is the protocol that stops it slowly. Open this next11 min- Diagnostics · 11 min Packet Loss vs Latency, and Why They Feel the Same to Users Every node in a ring is a hop, and hops are where the latency budget goes.
- Design · 9 min Hub and Spoke Topology, and the Traffic That Goes the Long Way The other topology chosen by arithmetic rather than preference, and the one most multi site businesses actually run.
- Design · 12 min Star Topology, and Why the Building Decided It The shape that won where the star lost.
- Infrastructure · 9 min Metro Ethernet, and What You Are Actually Buying From the Carrier Metro Ethernet, one of the places a ring is still the physical design.