Overview

Link Aggregation, defined by IEEE 802.3ad (Link Aggregation Control Protocol or LACP), combines multiple physical network ports into a single logical link. This technology increases total throughput beyond the limit of a single port and provides physical redundancy if a cable or interface fails. For B2B network architects and engineers, understanding LAG configurations, hardware compatibility, and performance limits is critical for core, distribution, or storage network design.

Frequently Asked Questions

Q1: What is the maximum switching throughput and backplane capacity for link aggregation on a 48-port Gigabit switch?

The maximum switching capacity for a standard 48-port Gigabit switch with four 10G uplinks is 176 Gbps (non-blocking). For link aggregation, a LAG of 8 x 1G ports yields a raw aggregate bandwidth of 8 Gbps full-duplex. Actual throughput depends on the hash algorithm; however, backplane capacity must be sufficient to support all active LAGs simultaneously. Always verify that the switch’s backplane is non-blocking (e.g., 176 Gbps for 48x1G + 4x10G) to achieve line-rate LAG performance.

Q2: Is link aggregation compatible with all SFP/SFP+ optical modules? Do I need identical transceivers?

Yes, link aggregation is compatible with most SFP, SFP+, and QSFP optical modules as long as all member ports within the same LAG use identical speed and duplex settings. Transceivers can be mixed between vendors (e.g., generic compatible and OEM) provided the port speed matches (e.g., all 10G-SR). However, for deterministic behavior and to avoid LACP negotiation failures, B2B deployments recommend identical module type, wavelength, and fiber mode across the LAG. Always check the switch vendor’s compatibility matrix before mixing third-party optics.

Q3: How does link aggregation differ from stacking? Which should I use for high availability?

Link aggregation (LAG) bonds ports across two or more switches to create a single logical link, increasing bandwidth and providing link-level redundancy. Stacking physically interconnects multiple switches (via dedicated stacking ports or cables) to operate as a single logical chassis with a single IP and management plane. For high availability: Use stacking for control-plane redundancy, simplified management, and cross-switch LAG (M-LAG or vPC). Use standard LAG for load balancing between two separate switches without stacking. For mission-critical core networks, combine both: stack two switches and then use LACP LAG to connect to downstream stacks.

Q4: How do I configure basic LACP link aggregation on a Cisco or Juniper switch (CLI example)?

On Cisco IOS, create a port-channel interface, set LACP mode to ‘active’, and add physical interfaces. Example:
interface port-channel 1
channel-group 1 mode active
switchport mode trunk
!
interface GigabitEthernet0/1
channel-group 1 mode active
!
interface GigabitEthernet0/2
channel-group 1 mode active
On Juniper JunOS, use:
set interfaces ae0 aggregated-ether-options lacp active
set interfaces ae0 unit 0 family ethernet-switching
set interfaces ge-0/0/0 ether-options 802.3ad ae0
set interfaces ge-0/0/1 ether-options 802.3ad ae0
Verify with ‘show etherchannel summary’ (Cisco) or ‘show lacp interfaces’ (Juniper).

Q5: What happens during a redundant power supply failover when link aggregation is active?

During a power supply failover (e.g., PSU1 fails, PSU2 takes over), link aggregation operation remains completely uninterrupted—traffic continues forwarding across all LAG member links. However, a failing power supply can induce voltage fluctuations or cause a partial switch reset if both PSUs are undersized. Ensure each PSU can independently handle full switch load (N+1 or 1+1 redundancy). After failover, LACP packets continue to be exchanged; the switch CPU remains powered, so link states do not flap. To troubleshoot, monitor ‘show power’ and ‘show lacp neighbor’ to confirm LACP remains active.

Q6: How does traffic load balancing work in a LAG and can I control it?

Link aggregation distributes frames across member links using a hash-based algorithm, not per-packet round-robin. Default hashing typically uses source/destination MAC addresses, IP addresses, or Layer 4 ports (configurable). For example, a hash of (src IP, dst IP) ensures all packets of a single flow use the same physical link, preserving packet order. You can modify the hash policy: On Cisco, use ‘port-channel load-balance src-dst-ip’. On Linux bonding, use ‘xmit_hash_policy layer3+4’. For maximum aggregate throughput, ensure traffic consists of many diverse flows. A single TCP connection between two hosts will never exceed the speed of one physical link in the LAG.

Q7: What is the maximum number of ports per LAG and does it vary by hardware generation?

IEEE 802.3ad allows up to 8 active ports per LAG, but many modern switches support 16 or 32 active ports (e.g., data center switches like Arista 7280R supports 32 ports per LAG). Older generation switches (pre-2010) often limited LAGs to 4 or 8 ports with lower backplane capacity. Next-generation switches (post-2018) support LACP with 128 member ports (e.g., some Broadcom Trident 3-based switches). Always check the switch’s datasheet for ‘Maximum number of ports per Link Aggregation Group’ and ‘Maximum number of LAG groups’ (typically 64-512 groups).

Q8: How do I troubleshoot a LAG that is not passing traffic or has flapping links?

First, verify LACP status: ‘show lacp neighbor’ should show partner system ID and operational state for each member port. Common causes: mismatched speed/duplex, mismatched VLAN trunk allowed lists, or one side configured as ‘static’ (on) while the other is ‘active’ or ‘passive’. Next, check cable integrity and SFP diagnostics (‘show interfaces transceiver’). Flapping links often indicate a bad SFP or exceeding the switch’s MAC address table limit. Lastly, verify that the hash algorithm is identical on both sides. For non-LACP static LAG, ensure both sides have identical member port configurations. Use ‘debug lacp’ (Cisco) or ‘traceoptions lacp’ (Juniper) cautiously in maintenance windows.