Using Link Aggregation in Your Network

IEEE 802.3ad link aggregation provides a standardized means of grouping multiple parallel Ethernet interfaces into a single logical Layer 2 link. The formed group of Ethernet interfaces is referred to as a Link Aggregation Group (LAG). Dynamic LAG formation and activation is provided by the Link Aggregation Control Protocol (LACP).

Each pair of LAG physical ports is made up of a local port on the device responsible for LACP negotiation, referred to as the actor, and its directly linked remote port on the device participating in the LACP negotiation, referred to as the partner. LAGs form automatically based upon a set of criteria (see How a LAG Forms).

Only LAG members in the attached state carry user traffic. Once the LAG is formed, the system ID, made up of a system priority and the device MAC address, determines which device will be in charge of choosing the LAG port members that will be moved to the attached state. While port speed is not a criteria for joining a LAG, the port speed must match for all ports that are placed in the LACP attached state. Aggregatable ports not selected to carry traffic for this LAG are available to the next LAG as long as LAG resources are not depleted. Should LAG resources become depleted, aggregatable ports are placed in LACP standby state.

802.3ad LACP aggregations can be run between combinations of switches, routers, and edge devices, such as a server, that support LACP.

Note

Earlier (proprietary) implementations of port aggregation referred to groups of aggregated ports as “trunks”.

The concept of grouping multiple ports into a single link is not a new idea. Cabletron's SmartTrunk, Cisco's Inter Switch Link trunking, and Adaptec's Duralink are previous examples. The problem with these older methods, from the network administrators point of view, is that they are proprietary. Administrators who wanted to implement faster logical links faced major problems if they also wanted, or needed, to use a different brand of networking hardware. Link aggregation is standards based allowing for interoperability between multiple vendors in the network.

Older implementations required manual configuration. With LACP, if a set of links can aggregate, they will aggregate. LACP‘s ability to automatically aggregate links represents a timesaver for the network administrator who will not be required to manually configure the aggregates. However, manual overrides are provided for when the administrator needs to customize. Link aggregation also provides for rapid configuration and reconfiguration when there are changes in the physical connections. Link aggregation will automatically and quickly converge the new configuration. This convergence typically occurs in one second or less.

Link aggregation is a cost effective way to implement increased bandwidth. A major benefit of link aggregation is the ability to incrementally add bandwidth in a linear fashion. Without link aggregation, if there is a need to increase the bandwidth for a 100Mbps pipe, the only choice is an exponential upgrade to a 1000Mbps pipe. If there is a need for a 300Mbps pipe, aggregating three 100Mbps ports is both less expensive, because a forklift hardware upgrade is avoided, and makes for more efficient use of the system ports that are already available.

The physical links within the aggregate can serve as redundant backups to one another. Since only a single MAC address representing the entire aggregate is presented to the MAC client, the failure of any link within the aggregate is transparent. Failover is handled within the link aggregation sublayer.