Link Aggregation Explained: A CompTIA Network+ Study Guide
Sooner or later, a single cable between two switches isn't enough. Traffic grows, users complain, and one failed port can take down a whole floor. Buying faster hardware is one fix. A cheaper, more resilient option is to bundle the links you already have. That's link aggregation, and you should expect it on the CompTIA Network+ exam, where it appears in the Network Implementation domain alongside other Ethernet switching features like VLANs, 802.1Q tagging, and spanning tree.
This guide covers what link aggregation is, how it works, the protocols behind it, and the details the exam likes to test.
What Is Link Aggregation?
Link aggregation combines multiple physical Ethernet links into a single logical link. Two switches connected by four 1 Gbps cables can treat those cables as one 4 Gbps connection. The switches, spanning tree, and higher-layer protocols all see one interface instead of four.
You'll hear the same idea under several names, and Network+ expects you to recognize them all. A link aggregation group (LAG) is the vendor-neutral term. Cisco calls it EtherChannel, and the logical interface is a port channel. On servers, the technique is usually called NIC teaming (Windows) or bonding (Linux). When a question uses any of these terms, it's describing the same basic concept.
Why Use Link Aggregation?
Link aggregation delivers two main benefits: more bandwidth and more redundancy.
Increased throughput. Bundling links raises the total capacity between two devices. This is especially useful on uplinks between access and distribution switches, or between a switch and a busy server, where many users share one path.
Redundancy and fault tolerance. If one cable or port in the bundle fails, traffic shifts to the remaining links. The logical link stays up, just with less capacity.
Better use of existing links. Without link aggregation, connecting two switches with multiple cables creates a loop, and Spanning Tree Protocol (STP) blocks all but one link to prevent it. Those extra cables just sit idle. Because STP treats a LAG as a single logical interface, all member links can forward traffic at once.
How Traffic Is Distributed
This is one of the most commonly misunderstood parts of link aggregation, and a favorite exam trap.
A LAG does not split a single data stream across all its links. Sending packets from one conversation down different cables could cause them to arrive out of order. Instead, the switch runs a hashing algorithm on header fields such as source and destination MAC addresses, IP addresses, or TCP/UDP port numbers. Each flow gets assigned to one member link, and every frame in that flow uses the same link.
The practical consequence is that a single flow can never go faster than one physical link. Four 1 Gbps links give you 4 Gbps of aggregate capacity, but one large file transfer between two hosts is still capped at 1 Gbps. The benefit shows up when many hosts and conversations share the bundle.
Distribution also isn't always even. If most traffic shares the same source and destination addresses, such as traffic between two routers, the hash may send nearly everything down one link. Changing the load-balancing method to include port numbers often spreads traffic more evenly.
Link Aggregation Protocols
There are three ways to form a LAG, and you should know how they differ.
| Method | Standard | Negotiation | Notes |
|---|---|---|---|
| LACP | IEEE 802.3ad / 802.1AX | Yes | Open standard, multivendor |
| PAgP | Cisco proprietary | Yes | Cisco-only environments |
| Static ("on") | None | No | No error checking |
LACP (Link Aggregation Control Protocol)
LACP is the open standard, originally defined in IEEE 802.3ad and later moved to IEEE 802.1AX. If an exam question mentions 802.3ad, think link aggregation. LACP devices exchange LACP data units (LACPDUs) to agree on which links belong in the bundle and to detect link failures or miswiring.
LACP ports run in one of two modes. An active port sends LACPDUs and tries to form a bundle. A passive port only responds to LACPDUs it receives. Active–active and active–passive pairings form a LAG, but passive–passive does not, because neither side starts the conversation. Expect a scenario question built on exactly that mistake.
A LACP bundle supports up to eight active links. Additional configured links can wait in standby and join if an active link fails.
PAgP (Port Aggregation Protocol)
PAgP is Cisco's proprietary alternative. Its modes are desirable (actively negotiates) and auto (waits to be asked). The logic mirrors LACP: auto–auto fails to form a bundle. Because it's Cisco-only, LACP is the better choice in mixed-vendor networks.
Static Aggregation
Static mode (Cisco's "on" setting) forces ports into a bundle without any negotiation. It works, but it gives up LACP's safety checks. If one side is set to "on" and the other isn't configured correctly, you can end up with forwarding problems or even a switching loop. Static mode also can't be combined with LACP or PAgP on the other end.
Configuration Requirements
For links to join the same LAG, their settings must match. Member ports generally need the same speed and duplex, the same switchport mode (access or trunk), the same access VLAN or allowed VLAN list and native VLAN, and the same LAG protocol and compatible mode on both ends.
A simple Cisco example looks like this:
interface range GigabitEthernet1/0/1 - 2
channel-group 1 mode active
!
interface Port-channel1
switchport mode trunk
Network+ won't expect you to memorize vendor syntax, but seeing it helps the concepts stick. Settings applied to the port-channel interface carry down to the member ports.
Multi-Chassis Link Aggregation
A standard LAG connects two single devices, which means the switch on either end is still a single point of failure. Multi-chassis link aggregation (MLAG) solves this by letting one device build a LAG to two separate switches that act as one logical partner. Vendors implement it through switch stacking, Cisco vPC, or similar technologies. For the exam, just know that it extends link aggregation's redundancy from the link level to the device level.
Troubleshooting Link Aggregation
When a LAG doesn't behave, work through these common causes.
The bundle never forms. Check the modes first. Passive–passive (LACP) or auto–auto (PAgP) won't negotiate. Also confirm both sides use the same protocol.
Some member ports are suspended or error-disabled. A port whose speed, duplex, or VLAN settings don't match the rest of the group will usually be kept out of the bundle. Compare member port configurations carefully.
Throughput is lower than expected. Remember the per-flow hashing rule. A single large flow uses only one link. If many flows still pile onto one link, review the load-balancing method.
Loops or MAC address flapping. These often point to a static "on" configuration on one side paired with an unbundled configuration on the other. Using LACP helps prevent this.
Exam Tips
Keep these points straight on test day. Link aggregation combines physical links into one logical link for bandwidth and redundancy. The open standard is LACP, IEEE 802.3ad/802.1AX, while PAgP is Cisco proprietary. LACP passive–passive and PAgP auto–auto won't form a bundle. Traffic is balanced per flow, not per packet, so one conversation is limited to one link's speed. STP treats the LAG as a single interface, so no member links are blocked. And member ports must have matching settings.
Practice Question
A network technician bundles four 1 Gbps uplinks between two switches using LACP. Users report that a single large file transfer between two servers never exceeds 1 Gbps. What is the most likely explanation?
A. One side of the LAG is configured in passive mode
B. The LAG's hashing algorithm places each flow on a single member link
C. Spanning Tree Protocol is blocking three of the links
D. The member ports have a duplex mismatch
Answer: B. Link aggregation load-balances per flow, so a single transfer uses one physical link. Option A would prevent the bundle from forming only if both sides were passive. Option C doesn't apply because STP sees the LAG as one link. Option D would typically suspend a port rather than cap a transfer at exactly 1 Gbps.
Wrapping Up
Link aggregation is a simple idea, but a few details trip people up: negotiation modes, per-flow load balancing, and matching port settings. Master those, and you'll be ready for any Network+ question on LACP, port channels, or NIC teaming, and for configuring the real thing on the job.
No comments:
Post a Comment