Learning Objectives

• Understand why link aggregation improves bandwidth and redundancy
• Explain how LACP negotiates an EtherChannel
• Calculate load-balancing outcomes across aggregated links

What is Link Aggregation?

Link aggregation combines multiple physical Ethernet links into a single logical link. On Cisco switches, this is called an EtherChannel. The IEEE standard is Link Aggregation Control Protocol (LACP), defined in 802.3ad (now 802.1AX).

Instead of managing four 1 Gbps links separately, you bundle them into one 4 Gbps logical link. The switch treats the bundle as a single interface for STP, VLAN configuration, and management.

Benefits

Link aggregation provides three things that individual links cannot. First, increased bandwidth — a bundle of four 10 Gbps links gives 40 Gbps of aggregate throughput. Second, link redundancy — if one physical link in the bundle fails, traffic is redistributed across the remaining links with no STP reconvergence. Third, load balancing — traffic is distributed across the member links using a hash algorithm, maximizing utilization of all links.

Unlike STP, which blocks redundant links, link aggregation treats the bundle as one logical interface. STP sees the bundle as a single link and will not block individual members.

LACP Negotiation

LACP uses BPDU-like frames called LACPDUs to negotiate and maintain the bundle. The negotiation goes through these stages:

1. Passive or Active — each side is configured as active (sends LACPDUs) or passive (waits for LACPDUs)
2. Detection — LACPDUs are exchanged, and switches agree on a common link-aggregation group
3. Aggregation — if both ends agree on parameters (same speed, duplex, VLAN config), the links are aggregated
4. Steady state — LACPDUs are sent periodically to verify link health

At least one side must be active for LACP negotiation to succeed. If both sides are passive, no LACPDUs are exchanged and the links stay separate.

Load Balancing Algorithms

LACP itself does not define the load-balancing algorithm — that is implementation-specific. The switch uses a hash of frame header fields to pick which physical link carries each frame. Common hash inputs include:

| Algorithm | Hash Input | Best For | |---|---|---| | src-mac | Source MAC only | Routers (same source MAC per interface) | | dst-mac | Destination MAC only | Servers with many clients | | src-dst-mac | Source + Destination MAC | General purpose — balanced for most traffic | | src-dst-ip | Source + Destination IP | Routed traffic — very good distribution | | src-dst-port | L4 source + destination port | Per-flow granularity (TCP/UDP sessions) |

The hash result is used as an index into the number of member links. This means the distribution is per-flow — all frames of a single TCP connection go to the same physical link (preserving packet ordering), but different flows are spread across links.

Configuration Requirements

All member ports in a link aggregation group must share identical configuration:

• Same speed and duplex setting
• Same VLAN mode (all access in same VLAN, or all trunk with same allowed VLANs and native VLAN)
• Same STP settings
• Same spanning tree port type

If any parameter differs, LACP will not aggregate the links. On Cisco, mismatched ports are placed in suspended state (if using PAgP) or simply excluded from the bundle (LACP).

Key Takeaways

• Link aggregation bundles multiple physical links into one logical link for bandwidth + redundancy
• LACP (802.1AX) negotiates the bundle through LACPDU exchange
• At least one side must be active for LACP to negotiate
• Load balancing uses a hash (MAC, IP, or port) and is per-flow, not per-packet
• All member ports must have identical configuration (speed, VLAN, STP)
• Link aggregation prevents individual links from being blocked by STP