Uncovering an AP Issue During Real Client Testing

While evaluating the Access Point's performance, we conducted basic throughput tests for TCP and UDP protocols in various directions (Downlink, Uplink, and Bidirectional) using over 70 real devices with per client load of 5Mbps for a duration of 10mins. During this process, there are some interesting observations identified.

We observed significant variations in channel utilization between the 2.4GHz and 5GHz bands under different network conditions. In the idle state with no clients connected, we observed low channel utilization, with 6% for the 2.4GHz band and just 1% for the 5GHz band. This performance remained consistent even during a ten-minute duration of no client connections. However, when we introduced a substantial load with more than 70 real devices connected, we saw a significant increase in channel utilization, with 42% on the 2.4GHz band and 54% on the 5GHz band.

Notably, while running the UDP Bi-Directional traffic with 70+ Real devices. Surprisingly, both the 2.4GHz and 5GHz bands showed channel utilization exceeding 90%, leading to an AP reboot.

Upon inspecting the Event logs, we noticed that the real clients connected to the AP were Deauthenticated and Disassociated.

Further examination via Packet Capture showed Deauthentication packets with "Reason code: Disassociated due to inactivity (0x0004)" which appears when a client device is disconnected (disassociated) from an access point due to perceived inactivity.

To pinpoint the exact occurrence of the reboot problem after running on multiple devices, we performed a series of tests by gradually increasing the device count. While using UDP Bidirectional traffic at varying per-client loads (5Mbps, 10Mbps, 40Mbps).?

Initially, we tested with 10 devices, and the AP operated smoothly. We then increased the device count to 20, and there were still no issues noted during the identical UDP Bidirectional traffic test.

However, when we conducted the test with 21Real devices and more, the channel utilization on both the 2.4GHz and 5GHz bands started to rise and exceeded 90%, regardless of whether we used high or lower per-client loads (5Mbps, 10Mbps, 40Mbps).

Therefore, our findings indicate that when running UDP Bi-directional traffic with 21 or more real devices, the Access Point's Channel Utilization significantly increases, ultimately resulting in an Access Point reboot.

Conducted a throughput test involving 70 virtual clients with per client load of 5Mbps for a duration of 10mins. And observed that there are no client disconnections throughout the test. This emphasizes the importance of real-world testing with Real devices to uncover and address interesting issues.