Field feedback indicates symmetrical 10G broadband speeds. However during speed testing on the S125 switch the downstream rate was found to be low. When testing on the upstream CR16K router both uplink and downlink speeds were Giga.
1. According to the on-site feedback configuration, it was found that Layer 2 aggregation was configured between S125 and the upstream CR16K router, with link-aggregation load-sharing mode flexible enabling per-packet load sharing. It is suspected that this configuration causes packets from the terminal to the server to be sent through different interfaces in the aggregation group, resulting in incorrect packet sequencing when they reach the upstream server. This leads to multiple TCP retransmissions and a decrease in average transmission rate.

2. Conduct an on-site local FTP speed test by uploading files from a computer under S125 to the server under CR16K. Compare this with uploading files from under CR16K to the FTP server. Uploading from under S125 takes over 3 minutes, while uploading from under CR16K takes only about 1 minute. Packet capture on the terminal verified a large number of TCP Out-Of-Order packets, confirming the above speculation.

Remove the link-aggregation load sharing mode flexible configuration and switch to per-flow load sharing. On-site speed test results are normal.