What Is a Network Hub?
Learn how Ethernet network hubs operate as Layer 1 multiport repeaters, why hubs create shared collision domains, and how switches improve LAN performance.
A network hub is a basic Ethernet device used to connect multiple devices on a LAN, or local area network. A LAN connects devices within a limited physical area, such as a home, office, classroom, or laboratory.
A hub operates at OSI Layer 1, the Physical layer. It does not understand MAC addresses, Ethernet frames, or IP addresses. Instead, it receives an electrical signal on one port, regenerates that signal, and repeats it through the other active ports.
Because of this behavior, a hub is also called a multiport repeater. It repeats physical signals but does not make intelligent forwarding decisions. Ethernet switches largely replaced hubs because switches can inspect Layer 2 frames and forward traffic more selectively.
What Does a Network Hub Do?
The main purpose of a hub is to provide a shared connection point for several Ethernet devices. For example, four computers can connect to a four-port hub, creating one shared Ethernet segment.
When a signal enters one hub port, the hub regenerates it and sends a copy through every other active port. The signal is not repeated back through the port on which it arrived.
- The hub receives the physical signal on the sending device's port.
- The hub regenerates the signal to help preserve its electrical quality.
- The hub repeats the signal through all other active ports.
- The hub does not identify the intended destination.
- Each receiving network interface decides whether the resulting Ethernet frame is relevant to its host.
This is physical-signal repetition, not frame forwarding. A hub does not read the destination MAC address and select an output port.
How Signal Forwarding Works
Consider four computers connected to a four-port hub: PC-A, PC-B, PC-C, and PC-D. PC-A sends a unicast Ethernet frame intended for PC-C.
- The electrical signal from PC-A enters the hub through PC-A's port.
- The hub regenerates the signal.
- The hub repeats the signal toward the ports connected to PC-B, PC-C, and PC-D.
- PC-C receives the signal and processes the frame because its MAC address is the destination.
- PC-B and PC-D also receive the signal, but their network interfaces normally discard the frame after checking its destination address.
The hub itself does not know that PC-C is the intended recipient. It simply repeats the physical signal to the other ports. This means every attached node sees transmissions, even when only one node should ultimately process the frame.
Network Hubs and the OSI Model
The OSI model is a framework for describing how network communication works in layers. The Physical layer, or OSI Layer 1, concerns the transmission of bits as physical signals over media such as copper cable, fiber, or wireless radio.
A hub belongs to Layer 1 because it works with the physical representation of data. It can regenerate and repeat electrical signals, but it does not interpret the Layer 2 data carried by those signals.
- Bits: Physical signals representing binary values.
- Ethernet frame: A Layer 2 data unit containing source and destination MAC addresses and other fields.
- MAC address: A Layer 2 hardware address identifying an Ethernet network interface.
- IP address: A Layer 3 logical address used by IP networking.
A hub does not learn a MAC address table. It does not compare a destination MAC address with a table, and it does not make a destination-based forwarding decision. For an overview of the layered model, see OSI Reference Model.
What Happens to Traffic on a Hub-Based LAN?
All devices connected to a hub share the same physical segment. Whenever one device transmits, the hub repeats the signal to all other active ports. This creates unnecessary traffic at ports that are not connected to the intended destination.
The receiving network interface examines the Ethernet frame. It normally accepts a frame when the destination is:
- The interface's own MAC address.
- An appropriate broadcast address, meaning the traffic is intended for every device in the relevant Layer 2 broadcast domain.
- A multicast address for a group to which the host belongs.
A host normally ignores a unicast frame addressed to another host. However, the signal has still reached that host's interface and consumed shared network capacity.
Broadcast and multicast traffic is expected to reach multiple devices. The main problem with a hub is that even ordinary unicast traffic is physically repeated to every other port.
Collision Domains and Shared Media
A collision domain is the portion of a network in which devices can experience collisions if they transmit at the same time.
Every port on a hub belongs to one shared collision domain. The connected devices share the same transmission medium and compete for access to it. If PC-A and PC-B transmit at nearly the same time, their signals interfere with one another and a collision occurs.
In shared Ethernet, devices detect the collision, stop transmitting, and retry after waiting according to the network's collision-handling behavior. This legacy process is associated with CSMA/CD, or Carrier Sense Multiple Access with Collision Detection.
As more devices use the hub, contention increases. All devices share the available bandwidth, and collisions or retransmissions can reduce effective throughput. A busy hub-based LAN therefore becomes less efficient as additional hosts and traffic are added.
One Hub, One Collision Domain
Imagine four PCs connected to one hub. PC-A, PC-B, PC-C, and PC-D are all in the same collision domain. A transmission by any one of them uses the shared medium and can conflict with a simultaneous transmission by any of the others.
This is different from a switched network, where each switch port normally represents a separate collision domain.
Half-Duplex Operation
Half-duplex communication allows an interface to transmit or receive at a given moment, but not both reliably at the same time. Ethernet hubs use half-duplex communication because all attached devices share one medium.
Half-duplex operation is necessary for collision detection. Devices must be able to listen to the shared medium, transmit when appropriate, and detect when another transmission interferes with theirs.
Full-duplex communication allows an interface to transmit and receive simultaneously. Modern switched Ethernet links normally operate at full duplex because each switch port provides a dedicated point-to-point link to a device. With no shared medium on that link, normal collision behavior is eliminated.
Security and Packet Capture Implications
Because a hub repeats signals to every other port, a device attached to the hub can potentially observe traffic sent between other devices. A network sniffer is a tool that captures and analyzes traffic visible to its network interface.
Historically, hubs were useful for protocol analysis and packet capture because an analyzer connected to an available hub port could receive copies of traffic from the shared segment. This made troubleshooting and observation straightforward.
The same behavior is a security weakness in a production network. A host should not normally be able to see the unicast traffic of unrelated hosts. Modern switched networks provide traffic separation by forwarding known unicast frames only toward the destination port. Network monitoring on a switch is normally designed deliberately, using features such as an authorized monitoring or mirror port rather than relying on accidental visibility.
Network Hub vs. Ethernet Switch
An Ethernet switch is typically a Layer 2 device that learns source MAC addresses and uses them to forward Ethernet frames. When a switch knows the destination MAC address, it generally sends a unicast frame only through the port associated with that address.
Same Four PCs: Hub and Switch
With a hub, PC-A's unicast transmission to PC-C is repeated toward PC-B, PC-C, and PC-D. All three receiving interfaces see the signal, and only PC-C normally processes the frame.
With a switch that has learned PC-C's MAC address, the switch normally sends PC-A's unicast frame only through PC-C's port. The other PCs do not receive that unicast frame from the switch. The links can also operate full duplex, avoiding the shared collision behavior of the hub.
Why Switches Replaced Hubs
Hubs are rarely deployed in contemporary Ethernet LANs. Switches provide several important advantages:
- They reduce unnecessary unicast traffic by forwarding known destinations selectively.
- They give each switch port its own collision domain.
- They normally support full-duplex communication.
- They reduce contention and improve effective throughput.
- They scale more effectively as the number of connected devices and traffic volume increase.
- They provide better traffic separation and reduce accidental exposure of unicast traffic.
Switches can still flood certain frames, such as broadcasts, multicasts depending on configuration, or unknown unicasts. However, this behavior is based on Ethernet frame handling and forwarding logic, not indiscriminate physical repetition of every signal by design.
Troubleshooting Hub-Based Networks
Poor Throughput When Several Hosts Are Active
Likely cause: All hosts share the hub's bandwidth and one collision domain. Increased contention and collisions reduce performance.
Diagnostic concept: Look for high collision rates and half-duplex operation on the shared Ethernet segment.
Resolution: Replace the hub with an Ethernet switch and use appropriate full-duplex links.
Duplex Mismatch or Excessive Late Collisions
Likely cause: A device may be configured for full duplex even though the hub can support only half-duplex communication.
Diagnostic concept: Verify the host interface's negotiated or configured duplex setting and account for the hub's half-duplex limitation.
Resolution: Use compatible half-duplex settings only as a temporary measure for legacy equipment, or migrate to a switch-based network. A duplex mismatch can cause poor performance, errors, and late collisions.
A Host Captures Other Hosts' Traffic
Likely cause: A hub repeats signals to all connected ports.
Diagnostic concept: Confirm whether the segment uses a physical hub rather than a switch or an intentionally configured monitoring port.
Resolution: Replace the hub with a switch and apply appropriate segmentation and monitoring controls.
Practical Summary
- A network hub is a multiport repeater.
- It operates at OSI Layer 1.
- It regenerates and repeats an incoming physical signal through every other active port.
- It does not inspect Ethernet frames or learn MAC addresses.
- All hub ports belong to one shared collision domain.
- Hub-connected Ethernet uses half duplex, so simultaneous transmissions can collide.
- Every connected interface sees repeated traffic, although the host normally discards frames not addressed to it.
- Switches replaced hubs because they provide selective Layer 2 forwarding, separate collision domains, full-duplex links, better scalability, and improved traffic separation.
Hubs are mainly important today as a foundation for understanding early Ethernet, shared media, collision domains, half-duplex operation, and the difference between Layer 1 signal handling and Layer 2 frame forwarding. For broader LAN concepts, see Computer Network Expained and Unicast Multicast Broadcast Addresses.