Typical case analysis: \"butterfly effect\" in wired networks.

Null

Wireless networks are built on the basis of wired networks, and the state of wired networks can directly affect the user experience of wireless networks. Network optimization is not limited to wireless network optimization, and the sorting and optimization of wired networks is sometimes more important.

Similar problems have been found in multiple locations:

Scenario 1: A certain network device is connected to a switch in the wired network, and the interface has not been configured as an edge port, resulting in frequent up-down phenomena.

Scenario 2: A certain AP in the network failed to register online due to a link or other reasons. According to the working mechanism of AP, it will automatically restart if it is not registered online after 10 minutes. Due to the lack of edge port configuration on the switch port, the interface connecting the switch and the AP will experience up-down phenomena.

Scenario 3: Dozens of APs in the network are not online for a long time due to link or license issues and restart every 10 minutes. Due to the lack of edge port configuration on the access switch, the interface connecting the switch and the AP will intermittently experience up-down phenomena, one after another.   

Once the up-down phenomenon occurs on a non-edge port of a switch, the switch will broadcast TC messages to the entire network, and various levels of switches will refresh MAC table entries. Especially in scenarios 1 and 3, the switch frequently broadcasts TC messages, and refreshing the MAC table entries will cause a large amount of unknown unicast flooding, greatly increasing the traffic in the network. Downstream devices will suffer from significant traffic shocks, leading to problems such as packet loss and lag, and severe cases may cause APs that are already online to drop offline.

A rough judgment method for unknown unicast flooding: the amount of traffic in the uplink direction of the switch is significantly less than the total amount of traffic in the downlink direction of all ports.

(1) In addition to checking the running status of AC and AP in the process of handling problems, engineers should also actively check whether there are unreasonable TC message records in the log buffer of aggregation switches and core switches, and suggest customers to rectify and eliminate the TC messages if any.

(2) Check whether the edge port is enabled on the access switch side. All switch interfaces that connect to APs need to be configured as edge ports.

(3) All downstream ports of switches below the core gateway should be configured with port isolation to some extent to improve the situation of unknown unicast flooding.