What Is ICMP Port: Detail Guide
The Internet Control Message Protocol (ICMP) is an integral protocol for communication across IP networks. ICMP is primarily…
Mesh topology connects each node to multiple other nodes, so there is more than one path between any two points. Lose a link and traffic reroutes around it. That redundancy is the entire reason mesh exists, and it is why mesh appears wherever an outage is unacceptable.
It is one of the five layouts covered in our network topology guide.
Full Mesh vs Partial Mesh
The distinction matters more than any other point on this page, because almost everything said about mesh is true of one and false of the other.
| Full mesh | Partial mesh | |
|---|---|---|
| Connections | Every node to every other node | Only important nodes are multiply connected |
| Links needed | n(n−1)/2 | Far fewer, by design |
| Hops between any two nodes | Always 1 | Sometimes more than 1 |
| Redundancy | Maximum | Good where it matters |
| Cost | Very high | Manageable |
| Scalability | Poor | Reasonable |
| Where used | Small critical cores | Most real WANs and fabrics |
In a full mesh, node A reaches node D over a direct link, never by hopping through B and C. Multi-hop forwarding is a partial mesh behaviour, and it is what wireless mesh networks do.
Why Full Mesh Does Not Scale
The link count is n(n−1)/2, which grows with the square of the node count:
| Nodes | Links required |
|---|---|
| 4 | 6 |
| 5 | 10 |
| 10 | 45 |
| 20 | 190 |
| 50 | 1,225 |
| 100 | 4,950 |
Every one of those links is a physical circuit, a port on both ends, and a routing adjacency to maintain. This is why full mesh is confined to small groups of critical nodes, and why partial mesh is what gets deployed in practice.
It is also one of the reasons SD-WAN exists, building a full mesh of MPLS circuits between dozens of branch sites is prohibitively expensive, so SD-WAN builds a mesh of tunnels over cheap internet links instead.
How Traffic Moves Through a Mesh
In a partial mesh, nodes forward on behalf of each other. If A has no direct link to D:
- A consults its routing table and picks the best next hop toward D, say B.
- B forwards toward C, C forwards to D.
- If the B–C link fails, the routing protocol reconverges and A’s traffic takes a different path.
That reconvergence is the whole point. Each extra hop adds latency, though, which is the trade-off against the redundancy. Protocols such as OSPF and BGP are what make path selection work at scale, and BFD is often added to detect failures in milliseconds rather than seconds.
Advantages of Mesh Topology
No single point of failure. With multiple paths, losing one link or node does not partition the network. This is the defining advantage.
Fast failure recovery. Alternative paths already exist, so recovery is a routing decision rather than a physical repair.
Traffic can be load-balanced. Multiple equal-cost paths let you spread load rather than forcing everything down one link.
Predictable latency in a full mesh. Every node is one hop from every other, so there are no long paths.
No central bottleneck. Unlike a star, there is no single device every packet must traverse.
Disadvantages of Mesh Topology
Cost. Each link needs cable or circuit, plus a port at both ends. Full mesh cost grows quadratically with node count.
Complexity to design and operate. More paths mean more routing state, more adjacencies and more that can be misconfigured. Diagnosing a problem means reasoning about which of several paths traffic actually took.
Poor scalability at full mesh. Adding one node to a 20-node full mesh means adding 20 new links.
Latency from multi-hop paths. In a partial mesh, some pairs of nodes are several hops apart, and each hop adds delay. This is a real constraint for wireless mesh in particular, where each hop also halves usable throughput on shared radio channels.
Redundancy is not security. Mesh improves availability, not confidentiality. Multiple paths do nothing to stop an attacker, that is the job of segmentation, ACLs and encryption.
Where Mesh Topology Is Actually Used
- WAN and service provider cores. Partial mesh between core routers, so no single circuit cut isolates a region.
- Data centre spine-leaf fabrics. Every leaf connects to every spine, a full mesh between the two layers, giving equal path length between any two servers.
- Wireless mesh networks. Access points relay for each other, so only some need a cable back to the switch. Ideal for warehouses, outdoor sites and older buildings where new cable runs are impractical.
- Home mesh Wi-Fi systems. The consumer version of the same idea, satellite nodes relay to a base unit to extend coverage. A wired backhaul between nodes avoids the per-hop throughput penalty.
- Industrial IoT and sensor networks. Protocols such as Zigbee and Thread use mesh so battery-powered sensors can relay for one another over short, low-power links.
- SD-WAN overlays. A logical mesh of encrypted tunnels between sites, built over ordinary internet connections.
Frequently Asked Questions
How many connections does a full mesh need?
n(n−1)/2, where n is the number of nodes. Ten nodes need 45 links, twenty need 190. The count grows with the square of the node count, which is what limits full mesh to small groups.
Is mesh topology scalable?
Full mesh is not, it is the least scalable common topology. Partial mesh scales reasonably because you choose where redundancy is worth paying for. When people say “mesh is scalable” they almost always mean partial mesh, or wireless mesh where nodes are added without new cabling.
Is mesh topology more secure than other topologies?
No. Mesh improves availability by removing single points of failure. It provides no confidentiality or access control on its own. Security comes from segmentation, authentication, ACLs and encryption regardless of topology.
What is the difference between mesh topology and mesh Wi-Fi?
Mesh topology is the general concept of multiply-connected nodes. Mesh Wi-Fi is a specific consumer product, several access points that relay wirelessly to extend coverage, which is a partial mesh. Each wireless hop roughly halves throughput unless the nodes have a wired or dedicated-radio backhaul.
Why does mesh Wi-Fi get slower with more hops?
A node using the same radio to receive and retransmit can only do one at a time, so usable throughput roughly halves per hop. Tri-band systems with a dedicated backhaul radio, or nodes wired with Ethernet, avoid this.
Is a spine-leaf data centre design a mesh?
Yes, a full mesh between the leaf and spine layers, though not within either layer. Every leaf connects to every spine, so any two servers are the same number of hops apart, which makes latency predictable.
Related Guides
Compare the alternatives in our network topology guide, or read about bus, star and ring topology. For the redundancy mechanisms behind mesh, see network redundancy and SD-WAN vs MPLS.
