A node in networking is any device on a network that can send, receive or forward data. Computers, printers and phones are end nodes. Switches, routers and firewalls are intermediary nodes. A patch panel has no address and is not a node at all.
- The test is a network connection and an address
- End nodes start or finish traffic. Intermediary nodes forward it
- Printers, cameras and access points are nodes, and they outnumber the computers
- A patch panel is not a node, however central it is to the cabling
- In a monitoring tool a node is a licensing unit, and definitions differ
On this page
The testWhat makes something a node
The test for a node in networking is simpler than most definitions make it sound. A device is a network node if it can send, receive or forward data across the network, which in practice means it has a network interface and an address.
That test excludes things people sometimes count by mistake. A patch panel is a set of physical connections and nothing more, and it does not appear on the network at all.
A media converter or an unmanaged hub moves signals between cables without addressing anything. Neither has an address and neither is a node, however important both are to the physical path.
It includes things people sometimes forget. Printers, cameras, access points, phones, badge readers, building management controllers and the uninterruptible power supply with a network card in it are all network nodes, and on real networks these devices frequently outnumber the computers.
The practical version of the question is rarely philosophical. Somebody counting network nodes is sizing a monitoring license, planning addresses across multiple networks, or drawing a diagram, and the right answer depends on which of those they are doing.
What they doWhat network nodes actually do
A definition explains what qualifies. It does not explain why the word is worth having, and that comes down to four things every network node does one or more of.
Communication. A node is a point at which communication can begin or end. Two computers exchanging data are two nodes and a path between them, and everything else on the network exists to make that path work.
Data transfer. Intermediary nodes move data across the network on behalf of others, deciding at each step which link the data leaves by. A packet crossing four networks passes through a chain of nodes, each of which makes that decision independently.
Resource sharing. Server nodes provide resources and services to the rest of the network: files, printing, authentication, databases, name resolution. This is what most end nodes are on the network to reach.
Reaching the internet. At the largest scale the same idea holds. The internet is nodes and links between them, and the routers inside a provider network are nodes exactly as a laptop is, doing the second job rather than the first.
The reason to keep those four in mind is that they explain the failure conversation. Asking which function a node was performing when something broke separates an end device problem from a data path problem, and those are different investigations.
Two typesThe two types of nodes that matter
Every classification of a node in networking reduces to one distinction, and the rest is detail.
End nodes originate data or consume it. A computer, a server, a printer, a phone, a sensor. Communication starts or finishes at these devices and no data passes through them on its way somewhere else.
Intermediary nodes forward data on behalf of other network nodes. Switches, routers, firewalls, load balancers, wireless access points. Communication passes through these devices, and their whole job is deciding where the data goes next.
Two things follow from that split, and they are the reason it is worth making.
Failure behaves differently. An end node that fails takes itself off the network and nothing else. An intermediary node that fails takes every device behind it, and the resources on the far side become unreachable, which is why network design spends its money on the second type and its inventory effort on the first.
Counting differs. A hundred-person office network has perhaps two hundred and fifty end nodes and fifteen intermediary ones. Monitoring tools bill for both types, and network diagrams normally draw the second and summarize the first.
The OSI world names these two types directly: an end system is a host, an intermediate system is a router, and the IS-IS protocol is called Intermediate System to Intermediate System for exactly that reason.
Where it narrowsWhere the word narrows
The same word means something more specific in four contexts, and confusion between these types is the usual source of an argument.
In a network topology diagram. A node is anything drawn at a connection point. In a star topology the nodes are the devices around the central switch, and in a ring network they are the devices in the loop. Here the word describes the shape of the network rather than what any device does.
In a monitoring or management tool. A node is a device the tool watches, and almost always a licensing unit. This is the meaning that costs money, and it is why "how many network nodes do we have" is more often a purchasing question than a technical one.
Read the vendor definition rather than assuming, because whether a switch counts as one node or as one per interface is a real difference across multiple networks.
In a cluster. A node is one computer that is a member of the cluster, whether that is a database cluster, a Kubernetes cluster or a hypervisor cluster. The nodes share resources and services within the cluster, and nothing about the network topology is implied by the word here.
In blockchain and peer to peer networks. A node is a participant holding a copy of the shared data. This is the meaning most likely to be intended by somebody who arrived at the word from outside IT operations.
None of those types is wrong. They are the same idea applied at different layers of a network, and the useful habit is asking which one is meant before answering a question about how many there are.
CountingCounting nodes, which is the question underneath
Most people asking what a network node is are about to count them, and the count is only as good as the boundary drawn around it.
For a monitoring license, count what the tool will discover on the network: every device with an address you want information about. Include the printers, the access points, the cameras and the power supplies. This number is always higher than people expect and it is the one with a price attached.
For addressing, count every device that needs an address in each subnet, then leave room. A subnet sized to today's node count is a network that needs renumbering in two years.
For a network diagram, draw the intermediary nodes and summarize the end nodes. Twelve switches and a note saying two hundred and forty endpoints is a readable diagram. Two hundred and fifty drawn devices is a poster nobody reads.
For a security review, the count that matters is network nodes you did not know about. Every network has some, and they are usually the building systems, the personal devices and the equipment a department bought without telling anybody.
PitfallsWhere people go wrong
Counting only the computers. Printers, phones, access points and building systems are network nodes too, and these devices usually outnumber the workstations. A monitoring quote based on staff headcount is wrong by a factor of two.
Assuming one vendor's node is another's. Licensing definitions differ, and a switch counted per chassis in one product is counted per port in another. Read the definition before comparing prices.
Treating a patch panel as a node. It has no address and does not appear on the network. It matters for the cable path and not for the count.
Confusing a node with a host. Every host is a node and not every node is a host. Switches and routers are network nodes and they are not hosts, which matters when a document says one type and means the other.
Drawing every node on a network diagram. The result is unreadable and it hides the structure. Draw the paths that carry the data, summarize the endpoints.
Forgetting the nodes nobody registered. The unmanaged switch under a desk and the camera system installed by a contractor are on the network whether or not they are in the inventory, and they are exactly the ones worth finding.
ComparisonFour words for overlapping things, and the row that separates them
| Criterion | Node | Host | Device | Endpoint |
|---|---|---|---|---|
| Sends or receives data | Yes | Yes | Sometimes | Yes |
| Forwards data for others | Sometimes | No | Sometimes | No |
| Includes switches and routers | Yes | No | Yes | No |
| Has a network address | Yes | Yes | Not always | Yes |
| Usual context | Topology and counting | Addressing and services | Anything physical | Security tooling |
| Includes a patch panel | No | No | Arguably | No |
The row that settles most arguments is the second. A host is something traffic starts or ends at, and a node is that plus everything in between. When a document uses the two words interchangeably it is usually describing end devices and has not thought about the switches.
FAQFrequently asked questions
What is a node in networking?
Any device on a computer network that can send, receive or forward data. Computers, servers, printers, phones, switches, routers and access points are all network nodes.
What are examples of network nodes?
A laptop, a network printer, a wireless access point, a switch, a router, or a server. Any device with a network interface and an address qualifies.
What is the difference between a node and a host?
Every host is a node and not every node is a host. A host originates or consumes traffic; a node includes the switches and routers that only forward it.
Is a router a node?
Yes. It is an intermediary node, because it forwards data for other devices on the network rather than originating any communication of its own.
Is a switch a node?
Yes, for the same reason. A managed switch has an address of its own as well, which makes it a node in every sense including the one your monitoring tool bills for.
Is a patch panel a node?
No. It is a set of physical connections with no address and no presence on the network, however central it is to the cabling.
What are the types of network nodes?
Two that matter. End nodes originate or consume data, such as computers and servers. Intermediary nodes forward that data for others, such as switches, routers and firewalls.
Why do monitoring tools count nodes?
Because it is their licensing unit. The definition varies between products, so a device counted once in one tool can be counted per interface in another.
How many nodes does a typical office network have?
Far more than the headcount. Printers, phones, access points, cameras and building systems commonly bring a hundred-person network to two hundred and fifty devices or more.
Is a node the same thing in a cluster?
No. There it means one server that is a member of the cluster, with no implication about network topology.
What is a node in blockchain?
A participant holding a copy of the shared data. It is the same word borrowed for a different kind of network, and a common source of confusion in a mixed conversation.
Does every network node need an IP address?
Every node that participates at the network layer does. A device operating only at layer two, such as an unmanaged switch, forwards frames without one and is a borderline case for exactly that reason.
What do network nodes do?
One or more of four things: begin or end communication, transfer data on behalf of others, provide resources and services to the rest of the network, or carry traffic within the internet itself.
How should I count nodes for a network diagram?
Draw the intermediary nodes and summarize the end nodes with a number. The structure is the useful part, and two hundred drawn icons hide it.
What is an intermediary device?
An intermediary device is a node that passes traffic along instead of creating or consuming it. Switches, routers, firewalls and wireless access points are intermediary devices. Computers, phones, printers and servers are end devices.
Keep readingRelated concepts
Read next · Addressing What Is a Subnet? Counting nodes per subnet is the other half of the question, and getting it wrong means renumbering later. Open this next15 min- Design · 12 min Star Topology, and Why the Building Decided It Where the word does most of its work: a topology is a description of how nodes are joined.
- Switching · 9 min The MAC Address Table, and How to Find Which Port a Device Is On What an intermediary node keeps so it can forward for everyone else, and where to look when it does not.