VMware ESXi and vSphere Cluster Management

Half-Duplex and Full-Duplex Ethernet

Learn how half-duplex and full-duplex Ethernet work, compare collisions and performance, configure auto-negotiation, and troubleshoot duplex mismatches.

Duplex describes how two connected devices exchange traffic. In networking, duplex communication means that data can travel in both directions between devices. However, communication in both directions does not always mean that both devices can transmit at the same time.

Ethernet interfaces use either half duplex or full duplex. Half duplex permits transmission in only one direction at a time. Full duplex permits simultaneous transmission and reception. Understanding this difference helps explain collisions, link performance, switch configuration, and difficult interface problems such as duplex mismatches.

This lesson focuses on Ethernet LAN links between network interface cards (NICs), switches, routers, and legacy hubs.

Duplex Communication Fundamentals

Duplex communication is communication in both directions between two devices. For example, a workstation can send an Ethernet frame to a switch, and the switch can send a frame back to the workstation.

There are two possible ways to provide those two directions:

• Half duplex: the interface transmits or receives at a particular moment, but it does not do both simultaneously.
• Full duplex: the interface transmits and receives simultaneously.

Duplex is an interface property of an Ethernet link. It is separate from speed, which is the link rate, such as 10 Mbps, 100 Mbps, or 1 Gbps. A link can therefore have a speed setting and a duplex setting, and both ends must use compatible values.

Half-Duplex Ethernet

A half-duplex Ethernet interface alternates between transmitting and receiving. It must wait for the shared medium to become available before transmitting. If another station transmits at the same time, the signals interfere and create a collision.

A collision is an event in which multiple devices on shared Ethernet media transmit simultaneously. Ethernet's traditional shared-media operation detects and handles collisions so stations can wait and retransmit. This behavior is associated with a collision domain, the portion of a network where simultaneous transmissions can interfere with one another.

Hubs and Half Duplex

A network hub is a Layer 1 shared-media device. It repeats received electrical signals out of its other ports rather than making independent forwarding decisions for each port. Every device attached to the hub shares the same collision domain.

A device connected through a traditional Ethernet hub must use half duplex. Only one station on the shared segment can successfully transmit at a time. If two stations transmit together, a collision can occur and the stations must recover using Ethernet's shared-media process.

Half duplex is uncommon in current LANs because modern networks normally use switches. A switch gives each connected interface a separate point-to-point link, removing the shared-media collision behavior found on a hub segment.

Full-Duplex Ethernet

A full-duplex Ethernet interface can transmit and receive at the same time. A workstation may send frames toward a switch while receiving frames from that switch on the same link.

Full-duplex Ethernet uses separate transmit and receive paths. In copper twisted-pair Ethernet, different wire pairs carry traffic in the two directions. Because the transmit and receive paths are separate, the endpoints do not contend for one shared transmission path.

On a properly operating full-duplex link, Ethernet collisions should not occur. Collision counters increasing on a healthy direct full-duplex link indicate a configuration, physical-layer, or equipment problem that requires investigation.

Common Full-Duplex Links

• Workstation NIC to a switch access port.
• Switch port to switch port.
• Router Ethernet interface to a switch port.
• Other direct Ethernet connections in which both endpoints support and agree on full duplex.

Both endpoints must support and agree on full-duplex operation. A switch configured for full duplex cannot make an incompatible endpoint operate correctly merely by changing its own setting.

Half Duplex Versus Full Duplex

Speed and duplex should always be considered separately. For example, “100 Mbps, full duplex” describes a 100 Mbps link that can transmit and receive simultaneously. “100 Mbps, half duplex” describes the same nominal rate but with alternating transmission and reception and possible collisions on shared media.

Ethernet Auto-Negotiation

Auto-negotiation is an Ethernet capability-exchange process. Connected interfaces advertise supported speeds and duplex modes, then select a mutually supported operating mode. On many modern links, auto-negotiation is the normal and preferred configuration.

For a direct switched Ethernet link, use automatic settings on both endpoints when supported. The interfaces can then select the highest mutually supported speed and full duplex. Afterward, verify the actual negotiated result rather than assuming that the intended setting took effect.

Manual configuration is appropriate only when there is a specific compatibility or operational reason. If one endpoint is manually configured, the corresponding endpoint must use compatible settings. Mismatched manual and automatic configuration can leave the link physically up while causing severe performance problems.

Expected Settings by Link Type

Duplex Mismatch

A duplex mismatch occurs when the two ends of an Ethernet connection use incompatible duplex modes. A common example is a switch port forced to full duplex while the connected NIC operates at half duplex.

The physical link may remain up because the devices can still detect a signal. That does not mean the link is operating correctly. The full-duplex side transmits without expecting collisions, while the half-duplex side uses shared-media rules and may detect collisions when traffic overlaps.

Common Symptoms

• Poor or inconsistent throughput.
• Intermittent connectivity or applications that appear to stall.
• Late collisions on the half-duplex side.
• Frame, input, receive, CRC, or frame-check-sequence errors.
• Excessive retransmissions by higher-layer protocols.
• A link that appears operational but performs badly under load.

A mismatch may affect traffic more noticeably in one direction. For example, the half-duplex endpoint may report collisions when it transmits at the same time as incoming traffic. The full-duplex endpoint may instead report receive errors or corrupted frames. Traffic volume and direction determine which counters increase most quickly.

Structured Duplex Troubleshooting

1. Identify the link. Determine which two interfaces are connected and whether the remote device is a switch, router, host, hub, or another shared-media device.
2. Inspect both endpoints. Record the operational speed and duplex, not only the administrative configuration. Check whether each side is set to auto, forced full, or forced half.
3. Review counters. Look for collisions and late collisions on a half-duplex interface. Also review input errors, receive errors, CRC errors, and other frame errors.
4. Check for a configuration mismatch. Pay particular attention to one side being forced while the other side is auto-negotiating.
5. Correct both sides. For a normal direct switched link, use auto-negotiation on both endpoints or explicitly configure matching full-duplex values. For a legacy hub connection, use matching half-duplex settings.
6. Verify after the change. Confirm the resulting speed and duplex on both endpoints, clear or note old counters if appropriate, and observe the interface under normal traffic.

Cisco IOS Inspection Commands

On a Cisco IOS switch, inspect the detailed state and counters of an interface with:

show interfaces gigabitEthernet 1/0/1
show interfaces status

Look for the displayed speed and duplex values. In the detailed output, review collision, late-collision, input-error, CRC, and related counters where the platform supports them. Compare these results with the connected device's interface status.

Restoring Auto-Negotiation

For a modern direct Ethernet connection, the switch interface can be configured to use automatic speed and duplex selection:

configure terminal
interface gigabitEthernet 1/0/1
speed auto
duplex auto
end

Apply compatible automatic settings on the remote endpoint as well, then verify the negotiated result. Auto-negotiation is normally preferable for modern switched links because it allows both interfaces to select a mutually supported mode.

Configuring a Legacy Hub Link

If a port must connect to a legacy hub, half duplex is expected. When a fixed setting is required, a Cisco IOS example is:

configure terminal
interface fastEthernet 0/1
speed 100
duplex half
end

Use this only when the connected device requires it. Configure the corresponding endpoint for the same speed and half-duplex mode. After changing the switch, verify the remote interface and confirm that the shared segment is intentional.

Practical Ethernet Scenarios

Legacy PC Through a Hub

A PC NIC connected to a traditional hub uses half duplex. The PC and every other station on the hub share one collision domain. Only one station can successfully transmit at a time, and simultaneous transmissions can collide.

Modern Workstation to Switch

A workstation connected directly to a managed access switch normally negotiates full duplex with the switch port. The workstation can send and receive simultaneously, and a healthy link should not report collisions.

Switch-to-Switch Uplink

A switch-to-switch Ethernet uplink should normally negotiate the highest mutually supported speed and full duplex. Traffic can travel in both directions at once, making full use of the dedicated link paths.

Manually Configured Mismatch

Suppose a switch is forced to full duplex while the endpoint uses automatic negotiation and resolves to half duplex. The link may show as up, but transfers can be slow. The half-duplex endpoint may report collisions or late collisions, while the full-duplex endpoint may report receive or frame-check-sequence errors.

Restore auto-negotiation on both sides when possible. Alternatively, configure identical speed and duplex values on both endpoints, then verify that errors and retransmissions stop increasing.

Design and Operational Guidance

• Use full duplex for direct Ethernet links between end devices and switches.
• Use full duplex for switch-to-switch links and router-to-switch Ethernet links.
• Use half duplex only when a legacy shared-media device, such as a hub, requires it.
• Prefer auto-negotiation on modern Ethernet links and use compatible settings at both ends.
• After changing an interface, verify the operational mode on both endpoints.
• Do not interpret a link light as proof that speed and duplex are correct.
• Remember that modern switched LANs generally eliminate shared-media collision behavior on direct switch links.

Related Ethernet Terms

A NIC is the network interface card or controller that connects a host to Ethernet. A switch forwards Ethernet frames between ports and commonly provides dedicated full-duplex links.

Ethernet traffic can be unicast, addressed to one destination interface; multicast, addressed to a defined group of receiving interfaces; or broadcast, addressed to every device in the local broadcast domain. These delivery types do not change the basic duplex rule: a shared hub segment requires half duplex, while a properly configured direct switched link normally uses full duplex.

Continue reviewing half- and full-duplex Ethernet concepts as part of your Ethernet switching and interface troubleshooting study.