An Ethernet hub and a network switch are networking devices that can both have a row of RJ45 ports, but they handle network traffic differently. In a computer network, a hub is a physical-layer repeater: a signal received on one hub port is repeated toward the other ports in the same shared network segment. A switch reads Ethernet frames, learns source MAC addresses, and forwards ordinary unicast frames toward the switch port where the destination was learned.
This source-backed comparison uses current standards and vendor documentation. No switch or hub was opened, benchmarked, or tested with live traffic for the article.
Network switch vs hub: key differences
| Trait | Ethernet hub | Network switch |
|---|---|---|
| Forwarding basis | Repeats received signals | Uses Ethernet frames and a MAC table |
| Traffic scope | Shared by ports in the segment | Known unicast sent toward one port |
| Duplex | Half-duplex shared operation | Full duplex on current point-to-point links |
| Collisions | One shared collision domain | No collisions on a correct full-duplex link |
| Current use | Legacy and special lab work | Normal wired LANs |
Both devices can act as a central connection point for multiple devices on the same Ethernet network, but their behavior is not interchangeable. A hub repeats incoming data across the shared medium. A switch makes a forwarding decision for each Ethernet frame, giving modern networks more usable capacity and more control.
A switch still forwards broadcast traffic, and it may flood a unicast frame whose destination has not been learned. That does not make it a hub. Once the destination address is in its MAC address table, the switch can send later unicast network traffic only toward the relevant port.
Difference between a hub and switch in the OSI model
A network hub works at the physical layer, or Layer 1, of the OSI model. It repeats electrical or optical signaling without making a decision about the destination device. An Ethernet switch normally works at the data link layer, or Layer 2. It reads Ethernet frames and uses MAC addresses for frame forwarding and filtering within a local area network, or LAN.
People often say that both devices move data packets, but the Layer 2 switch is forwarding frames on the local network. A router works at the network layer with IP packets. Keeping those layers separate makes the hub vs switch vs router comparison easier to understand.
Collision domains and broadcast domains
All ports on a classic Ethernet hub share one collision domain. Each full-duplex switch port is its own collision-free point-to-point link. A basic switch does not split the broadcast domain by itself: connected devices in the same VLAN still receive broadcasts. A router or Layer 3 policy separates IP broadcast domains.
This is the key difference for multiple devices. A network hub makes every device share the same segment and bandwidth. A network switch gives connected devices separate access links with dedicated bandwidth at the edge, subject to uplink and switching-fabric limits, and forwards ordinary unicast data toward a specific device.
How Ethernet traffic moves
What is an Ethernet hub?
A classic hub does not build a list of computers or inspect an IP address. It regenerates the physical signal across the shared network segment. Every attached network interface receives that traffic and decides whether to accept the frame. All hub ports and connected computers share the available bandwidth.
Traditional shared Ethernet uses half-duplex mode and carrier sense multiple access with collision detection. Only one device should transmit at a time. If two transmit together, the signals collide, the devices stop, wait, and try again. More active devices raise contention and reduce useful capacity.
What is an Ethernet switch?
A Layer 2 switch reads the source MAC address of an arriving frame and associates it with the incoming port in a switching table. When the destination MAC address is known, the switch forwards the frame toward that specific port. The MAC table ages over time and changes when a device moves.
Broadcast traffic still reaches other ports in the same VLAN. Multicast handling depends on the switch and its settings. An unknown unicast may be flooded until the destination is learned. A managed switch can add VLANs, access rules, port monitoring, and multicast controls, while an unmanaged switch handles the basic forwarding automatically.
Full duplex changes the link
A current switch and endpoint normally negotiate a full-duplex point-to-point link. Both sides can transmit at the same time, and CSMA/CD collisions do not apply. Each access port has its own link rate rather than sharing one half-duplex segment with every port.
A duplex mismatch can still cause severe slowness on older equipment. One side may operate in full duplex while the other uses half duplex, producing errors, loss, and retries. Leave autonegotiation enabled on both sides unless the equipment documentation requires a matched fixed setting.
Types of Ethernet switches
An unmanaged Ethernet switch provides basic frame forwarding with no configuration interface. It is the normal choice for adding a few wired ports to a trusted home network. A managed network switch adds network management settings, status, logs, VLANs, traffic controls, and monitoring. A smart switch usually offers a smaller managed feature set.
A Power over Ethernet switch can deliver data and power to supported cameras, access points, and phones. Port count alone is not enough: the switch also needs the right PoE type, total power budget, per-port power, and cooling. A Layer 3 switch can route between IP networks, but its policy and firewall role should be defined rather than assumed from the product label.
Network performance, speed, and capacity
A 100 Mb/s hub does not give 100 Mb/s to every device at once. The segment shares that capacity and only one station transmits at a time. A gigabit switch gives each gigabit endpoint its own full-duplex access link for data transfer, though traffic can still meet congestion at an uplink, server, internet connection, or internal switch limit.
“Switching hub” is a confusing marketing term that commonly describes a switch. A product that supports full-duplex Ethernet and learns addresses is not a classic shared repeater. Many dual-speed hubs also contain internal switching between speed groups, which makes them poor tools for packet capture.
Backplane capacity and latency
A switch has to move traffic between ports through its internal fabric. The switching capacity or backplane figure helps show whether many ports can run near line rate together. Packet forwarding rate, uplink bandwidth, and buffer design also matter. A small office may never stress these limits, while a storage or camera network can.
Latency through a current switch is normally small, but the exact result depends on hardware, packet size, queueing, congestion, and features. A hub begins repeating a signal without making a forwarding decision, yet its shared half-duplex segment creates collisions and contention. That does not make an Ethernet hub the faster production choice.
Security and monitoring
On a hub, frames are repeated across the shared network, so a network interface in promiscuous mode may see traffic between other devices. On a switched LAN, ordinary known unicast traffic between two other ports is not normally delivered to a third port. Encryption is still required because switching and its security features do not create a privacy boundary.
A managed switch can mirror traffic from selected ports or VLANs to an analysis port. A network tap can provide a dedicated capture path. These methods are more controlled than inserting an old hub, which forces half-duplex behavior, can limit speed, and may drop capture traffic under load.
Capture only networks and systems you are authorized to inspect. Record the capture point, direction, speed, and any packet loss. A mirrored port can be oversubscribed when traffic from several full-duplex links is copied to one analyzer connection.
Network switch traffic management
VLANs divide one physical switch into separate Layer 2 networks. They can separate guests, cameras, phones, servers, and management traffic when the router and access policy are configured to match. A VLAN is not useful isolation when every network is routed together without restrictions.
Quality of Service, or QoS, marks or queues selected traffic when a link is congested. It does not create bandwidth. Spanning Tree Protocol helps block redundant Layer 2 paths that would otherwise form a switching loop. A loop can flood broadcasts and make the LAN unusable within seconds.
Port security, MAC limits, disabled unused ports, protected management access, and current firmware can reduce avoidable exposure. These controls are reasons to choose a managed network switch for a business. They also create configuration work, so keep a backup and document the intended port, VLAN, and uplink settings.
When to use a network switch or Ethernet hub
Home and small office
Use an unmanaged gigabit switch when the job is adding ports to a trusted network. Pick a managed switch when guest devices, cameras, phones, or lab systems need VLAN separation or monitoring. An Ethernet hub adds no useful benefit to a normal home LAN.
Business and production
Use managed switches with the port speed, uplinks, Power over Ethernet, security features, support, and monitoring the design requires. Business networks benefit from port labels, configuration backups, firmware policy, and logs. A hub hides device-level port status and creates a shared failure domain.
Industrial and legacy equipment
Some old controllers, test fixtures, or training systems were designed around 10BASE-T or half-duplex behavior. Do not replace a working legacy segment blindly. Check supported speed, duplex, pin use, timing, electrical environment, and vendor instructions. An industrial switch with fixed 10 Mb/s and half-duplex support may provide a cleaner migration path.
Packet-analysis lab
An actual hub can expose shared traffic for a small controlled lab. It may also change the behavior being observed by forcing half duplex and a lower speed. Port mirroring, a tap, or capture on the endpoint usually preserves a more realistic switched link.
Hub vs switch vs router
A basic Layer 2 switch connects devices in one Ethernet LAN or VLAN. A router is a different networking device that moves IP packets between computer networks, such as a home LAN and the internet. A typical home router combines routing, firewall, Wi-Fi, and a small internal Ethernet switch in one enclosure.
Adding a switch expands the wired side of that router and provides more network connections for local network equipment. It does not create another internet connection or protect untrusted devices by itself. For separate networks and IP addresses, the router or Layer 3 switch needs interfaces and policy for those networks, and the access switch must place ports in the intended VLANs.
Buying guide: choosing an Ethernet switch instead of a hub
- Count active devices and growth. Leave ports for an uplink, access points, cameras, phones, and likely near-term additions.
- Choose link speeds. Match endpoint needs, server uplinks, cabling, and the router rather than paying for speed no device can use.
- Set the PoE requirement. Add device power needs, startup margin, port count, and the required PoE type.
- Choose management depth. Use unmanaged switching for trusted port expansion and managed switching for VLANs, QoS, monitoring, and controlled redundancy.
- Check physical fit. Consider desktop or rack mounting, fan noise, power supply, operating temperature, and cable direction.
- Review support. Look for firmware history, published security notices, warranty terms, configuration export, and replaceable power supplies where downtime matters.
A used enterprise switch can offer many features at a low purchase price, but it may be loud, power-hungry, out of support, or locked behind an account. A current unmanaged switch can be the safer home choice. Compare lifetime power, adapters, fans, firmware, and replacement access as well as port price.
Installation: how to replace an Ethernet hub
- Record each cable. Label the device and hub port before disconnecting anything.
- Check endpoint support. Record current link speed and duplex for old equipment.
- Select the switch. Count ports, uplinks, PoE needs, mounting, temperature, and management needs.
- Use known-good cabling. Replace damaged cords and test fixed runs if errors are present.
- Connect the uplink once. Avoid accidental loops between the new switch and the rest of the LAN.
- Move devices in small groups. Confirm link, address assignment, gateway access, and application traffic.
- Watch errors. Check negotiation and error counters when an old endpoint becomes slow.
A direct switch replacement normally needs no IP-address change. The endpoints keep using the same router and subnet. A managed replacement may arrive with every port in one default VLAN, but its management account and firmware still need attention before production use.
Network switch vs hub questions
Do I need a hub or switch?
Use a network switch for normal data transfer between multiple devices. Use an Ethernet hub only when a controlled lab or legacy system specifically needs shared half-duplex behavior.
Does a switch send data to every port?
It sends broadcasts and some multicast or unknown traffic more widely within the VLAN. Known unicast traffic is normally forwarded toward the learned destination port.
Can a switch have collisions?
A correct full-duplex point-to-point Ethernet link does not use collision detection. Collision counters on a current LAN can point to half-duplex operation, a duplex mismatch, old equipment, or a reporting detail that needs investigation.
Is a network switch always better than a hub?
For a current home or business network, yes. A real hub still has a narrow use in training, legacy compatibility, or an authorized packet-capture lab. Those cases depend on its shared-media behavior, not better everyday performance.
What are the downsides of a managed switch?
It costs more, takes time to configure, and can break traffic through a bad VLAN, loop, access, or uplink setting. Use the smallest feature set that meets the network plan, save the configuration, and keep a recovery path.
Can an unmanaged switch act like an Ethernet hub?
No. It still learns MAC addresses and forwards known unicast frames toward the learned port. It lacks a management interface, but its basic traffic handling remains switching.
Sources
- Cisco explanation of switching and hubs, checked July 16, 2026.
- IEEE 802.3 Ethernet Working Group, checked July 16, 2026.
- Wireshark Ethernet capture reference, checked July 16, 2026.