The use of packet testing software can lead to significant discrepancies in upstream and downstream testing when evaluating APs.

AC——SW——AP.

During local forwarding speed tests, the upstream and downstream speeds are generally symmetrical. However, during centralized forwarding speed tests, there can be large differences in upstream and downstream speeds measured on the AP.

After analysis, it was found that the packet testing software negotiated a large TCP window size during testing, causing packets to be fragmented. The upstream CapWAP packets were fragmented on the AC because the FPGA did not have the ability to reassemble them, resulting in lower performance when using AC software forwarding. The downstream speed test showed normal performance because fragmentation did not occur on the AC. In normal Internet usage, software generally does not negotiate a very large TCP window size. Even if it does, not every packet would be large enough to require fragmentation, so it would not affect normal browsing speeds.

Configuring the TCP MSS (Maximum Segment Size) can solve this issue.