What is link aggregation using LACP, and how does it improve both bandwidth and reliability? What are the common issues that can occur with misconfigured LAG bundles?

Question

Accepted Answer

Link aggregation groups multiple physical links into a single logical link, providing both increased bandwidth and redundancy. The Link Aggregation Control Protocol, LACP, defined in IEEE 802.3ad (now part of 802.1AX), is the standard protocol for negotiating and maintaining link aggregation bundles between network devices.

LACP works by exchanging control frames, called LACPDUs, between the two devices on each physical link. These frames carry information about the system identity, port priority, and aggregation capabilities. When both sides agree that their ports should be bundled, the links form a port channel or LAG. LACP can operate in active mode, where the device initiates LACP negotiation by sending LACPDUs, or passive mode, where the device only responds to LACPDUs from the partner. At least one side must be in active mode for negotiation to succeed.

The bandwidth benefit comes from distributing traffic across all member links. A four-member LAG of 10 Gbps links provides 40 Gbps of aggregate bandwidth. However, it is important to understand that individual flows do not span multiple links. Traffic is distributed across links using a hashing algorithm based on source and destination MAC addresses, IP addresses, or port numbers. Each flow is assigned to a specific member link. This means that a single TCP connection between two hosts will use only one member link's bandwidth. The aggregate bandwidth is only realized when multiple flows traverse the LAG simultaneously.

The reliability benefit is that if one member link fails, traffic redistributes across the remaining links. The LAG continues to operate with reduced aggregate bandwidth but without any connectivity interruption. LACP detects member link failures through the absence of LACPDUs and removes the failed link from the bundle. The minimum and maximum number of links can be configured so that the LAG only operates when a sufficient number of members are active.

Common misconfiguration issues include several problematic scenarios. Speed and duplex mismatches between member links cause LACP to reject the mismatched link, resulting in fewer active members than expected. Inconsistent VLAN configurations across member links create a situation where traffic on some VLANs uses all links while other VLANs only use a subset, leading to unpredictable load distribution. LACP mode mismatches, where both sides are in passive mode, prevent the LAG from forming because neither side initiates negotiation. Connecting LAG members to different switches without using MC-LAG or VSS creates a split LAG where spanning tree blocks links, defeating the purpose of aggregation.

Hashing algorithm selection significantly affects traffic distribution. If most traffic is between two specific IP addresses, IP-based hashing assigns all that traffic to a single link regardless of how many links are in the bundle. Selecting a hash algorithm that includes Layer 4 port numbers provides better distribution when traffic is concentrated between a few hosts but uses many different application ports. Monitoring per-member link utilization helps identify uneven distribution that wastes capacity on underutilized members.

LACP fast timers should be configured for environments where rapid failure detection is important. The default LACP timeout is 90 seconds, meaning a failed link may not be removed from the bundle for up to 90 seconds. Fast timers reduce this to 3 seconds, providing much faster convergence. However, fast timers increase control plane overhead, which is typically negligible on modern hardware.