Port-Based Load Sharing

A port-based load sharing key is used to determine the index of the selected aggregator port where the list of aggregator ports is organized as an array sorted by increasing front panel port number with a zero-based index. The index of the selected aggregator port is equal to the key value module of the number of ports in the aggregator.
Note

Note

This feature is supported on all ExtremeSwitching Universal platforms.

index = key value % N

where N = the number of ports in the aggregator

The resulting behavior is that ports with a key value of 0 distribute to the lowest numbered port in an aggregator, ports with a key value of 1 distribute to the second lowest numbered port in an aggregator, etc.

Example

A port-based load sharing group contains aggregator ports 2,4,6 and 8. If the zero-based load sharing key 7 is assigned to port 1, then traffic received on port 1 and forwarded to the group will be transmitted on port 8 according to the following calculation:

7 (key) modulo 4 (number of ports in the aggregator) = 3 (index), which corresponds to port 8 which has zero-based index 3 in the sorted array of aggregator ports as shown in Aggregator Ports Array:

Table 1. Aggregator Ports Array
Index Keys Member port
0 0, 4, 8, 12 2
1 1, 5, 9, 13 4
2 2, 6, 10, 14 6
3 3, 7, 11, 15 8

Port-based Load Sharing Limitations

When considering the selection of a LAG algorithm, even distribution over member ports is usually the goal. Full utilization of the LAG‘s bandwidth requires even distribution. In the absence of even distribution, a single member port may become oversubscribed while other member ports are undersubscribed resulting in traffic loss when the LAG, viewed as an aggregate of member ports, is undersubscribed. LAGs distribute best when the diversity of flows destined for the LAG is large relative to the number of ports in the aggregator. For example, when many thousands of L2 flows are destined to a LAG using the “L2” algorithm, distribution on the LAG is typically good (even). Since the number of ports which will switch to a LAG is unlikely to be much larger (orders of magnitude) than the number of ports in the aggregator, extra care may be required from a network administrator when configuring and/or provisioning a switch using port-based LAGs.