Configure Congestion Avoidance

The congestion avoidance technology monitors the network resource utility and communication load of the network, so as to avoid the congestion by actively dropping packets before the network congestion happens or worsens. Excessive congestion exerts great harm on network resources and therefore, a certain measure must be adopted to relieve the congestion. The common measure is to configure the packet drop mode.

Configuration Condition

None

Configure Drop Mode

Tail drop and WRED (Weighted Random Early Detection) are two common packet drop modes.

Tail drop: It is a traditional packet drop policy. When the queue length reaches the maximum value, all new packets will be dropped. This packet drop policy may cause TCP global synchronization. When the packets connected by multiple TCPs are dropped in a queue, multiple TCP connections will enter the congestion avoidance and slow-start status to decrease and adjust the traffic, and then the traffic peak may occur simultaneously at a time. Repeatedly, the traffic and network are unstable.

WRED: When the queue length exceeds its own length, the packets are dropped by 100%. When the queue length is less than the start-value, do not drop any packet. When the queue length is greater than the start-value, drop packets at random according to the configured value. The random number generated by the WRED is based on the priority. The WRED introduces the IP priority to distinguish from the packet drop policy. The packet with high priority is considered for its benefit and this packet will be dropped in a relatively low probability.

Table 1-14 Configure the packet drop mode