LAG Overview


Link aggregation is a method of grouping multiple Ethernet ports into a virtual link with aggregated bandwidth. The system treats the set of ports in a link aggregation group (LAG) as a single port. All the ports within the LAG use the same Layer 2 MAC address and same default virtual LAN (VLAN). Traffic is distributed across the LAG in a way that ensures that traffic for a particular user stays in order. You can specify which port within the LAG to use as the flood port for broadcast or multicast traffic.

Within the system's interface and port hierarchy, a LAG can connect to the following upper layers: A LAG can connect to the following lower layers: LAGs support the following:

Using weights for traffic distribution across a LAG

Traffic is distributed over the LAG based on the weight set for each port in the configuration. The weight is set with the weight argument of the interface command. The value that you assign as a weight is relative to the weights of the other ports in the LAG. Any given port in the LAG carries a fraction of the entire LAG traffic that is equal to its weight divided by the sum of all weights of all ports in the LAG. For example, if you have a LAG with one 100M port and one 1000M port, you can configure the weighted distribution to be 10% for the 100M port and 90% for the 1000M port by specifying weights of 1 and 10, respectively. As the system receives packets, it hashes information in each received packet to an 8-bit value. (For Layer 2 traffic, the hash is based on MAC destination address and source address; for Layer 3 traffic, it is based on the IP destination address and source address.) The system uses this value, in conjunction with the configured weight, to select the port that will carry the traffic. In this way, all traffic that belongs to a given flow will always be forwarded across the same port in the LAG (and therefore is kept in order). If a port fails, the weight for each active port is regenerated based on the remaining active ports in the LAG. Traffic is then redistributed over all active links based on the new weights. For example, consider that you have three ports, weighted as follows:

Port Weight
port1 5
port2 10
port3 15

Flood ports on a LAG

LAG interfaces use a flood port to broadcast address requests for the switch and to flood packets belonging to unknown addresses, and for other broadcast and multicast traffic. Traffic whose destination address has not yet been learned is "flooded" out the flood port in an attempt to find the destination. You can specify the flood port preference with the interface command. When you configure an Ethernet port as part of the LAG, you assign a preference for selection as the flood port. The active port with the lowest value is selected as the flood port. The system displays the port preference value with the show interface command and displays the active flood port with the show command. If the port becomes unavailable, the software selects another flood port based on the flood port assignments (or defaults). When the original flood port again becomes active, it resumes as the flood port assuming it still has the highest ranking (lowest configured flood port preference). In the event of a tie in ranking, the system selects the port that first became active on the LAG.

Port configuration priority

You can configure characteristics of Ethernet ports through several mechanisms: