What link aggregation is
Link aggregation combines several compatible Ethernet links into one logical interface, often called a Link Aggregation Group (LAG), port-channel, bond, or EtherChannel depending on the platform. The goal is to increase aggregate capacity and provide redundancy without presenting each physical member as an independent Layer 2 path.
From the point of view of VLANs, trunks, and spanning tree, the bundle can be treated as one logical link while traffic is distributed across the physical members according to an implementation-specific hashing policy.
Why use a LAG
A LAG offers two main benefits. First, several flows can use more total bandwidth than one physical interface could provide. Second, losing one member does not necessarily bring down the logical connection if other healthy members remain.
This is especially useful on switch uplinks, server connections, storage paths, hypervisors, and other places where one link could become a bottleneck or single point of failure.
Aggregate capacity is not the same as one faster flow
Link aggregation usually distributes traffic using a hash based on fields such as MAC addresses, IP addresses, or transport-layer ports. One individual conversation is commonly kept on one member to avoid reordering.
Therefore, four 1 Gb/s links can provide roughly 4 Gb/s of aggregate opportunity across many suitable flows, but one single flow should not be assumed to reach 4 Gb/s. The benefit is strongest when traffic contains enough independent conversations for the hash to spread them across members.
LACP: standards-based negotiation
Link Aggregation Control Protocol (LACP), standardized in IEEE 802.1AX, lets devices negotiate and maintain an aggregation relationship. It helps verify that candidate links connect to the expected partner and can participate in the same logical bundle.
LACP is not the only way some vendors can create a static bundle, but dynamic negotiation improves operational visibility and can reduce the risk of silently forwarding over an inconsistent configuration.
LACP active and passive modes
In common terminology, an active LACP participant initiates protocol exchanges, while a passive participant waits for the other side. At least one side normally needs to be active for negotiation to begin.
The exact CLI depends on the platform, but the conceptual requirement is consistent: both ends must agree on the logical aggregation and the physical members must be compatible.
Member compatibility
Physical links placed in the same LAG should have compatible operational characteristics. Platforms commonly require matching speed, duplex, VLAN mode, trunking parameters, and other interface properties so all members can behave as one logical port.
If one link is configured differently, it may be suspended, excluded, or create unexpected behavior. When troubleshooting a partial bundle, compare the configuration and state of every member rather than inspecting only the logical interface.
Using a LAG as an 802.1Q trunk
A logical aggregate can itself operate as an 802.1Q trunk. In that case, VLAN tagging and allowed-VLAN policy belong conceptually to the logical bundle, while several physical links provide the underlying capacity.
Keep member configuration consistent with the bundle model. Applying conflicting VLAN or trunk settings to individual members is a common source of suspended links or traffic that behaves differently from the rest of the port-channel.
How aggregation interacts with spanning tree
STP/RSTP sees the aggregation as one logical Layer 2 link instead of several independent parallel links. This lets multiple physical interfaces carry active traffic without spanning tree blocking them merely because they connect the same Layer 2 devices.
Redundancy can then exist at two levels: member-level redundancy inside the LAG and path-level redundancy between different logical links controlled by spanning tree or another topology design.
What happens when one member fails
If one physical member fails, the bundle can remain operational with reduced capacity as long as enough valid members stay active. The hashing process is recalculated or remapped so traffic uses the surviving links.
This failover is useful, but it is not invisible to every application. A traffic flow may pause briefly, and the remaining links can become congested if the bundle was already heavily utilized before the failure.
Multichassis link aggregation
Traditional LACP aggregation expects one logical system at each end. Many vendors also provide multichassis technologies that let a downstream device form a logical bundle toward two physical switches. These designs can remove one-switch failure as a dependency for the attached host or access switch.
Implementation details vary significantly, so do not assume one vendor's MLAG, vPC, VSS, stacking, or chassis virtualization behaves exactly like another. The common objective is to present coordinated forwarding while preserving redundancy across devices.
Traffic distribution and hashing
Load distribution is normally deterministic rather than packet-by-packet random. A hash chooses a member based on selected header fields. This preserves frame ordering inside a flow while spreading different flows across the bundle.
A badly distributed traffic mix can leave one member busy while others are comparatively idle. If aggregate utilization looks uneven, inspect the hash policy and the diversity of source/destination addresses and transport sessions before assuming the bundle is malfunctioning.
How to troubleshoot EtherChannel and LACP
Check the logical interface first, then every physical member. Verify partner identity, LACP state, speed/duplex, VLAN/trunk configuration, and whether a member has been suspended or excluded.
- Do both ends place the same links in the same logical bundle?
- Is at least one LACP side active when dynamic negotiation is expected?
- Are member characteristics compatible?
- Does the logical trunk carry the expected VLANs?
- Is spanning tree operating on the logical interface as intended?
- Is traffic distribution limited by the flow mix rather than total bundle capacity?
What to study after link aggregation
Complete the access layer with Wi-Fi and inter-VLAN routing. At that point you can explain how endpoints attach, how VLANs segment the LAN, how trunks transport those VLANs, how STP controls redundant paths, and how LACP combines compatible links into one logical connection.