Hello,
i finally have my Windows game pc in its own subnet (malware/ransomware isolation precautions). I can reach my LAN (which for the OPNsense is WAN) and have actively disabled outbound-NAT for known source/destination subnets. The hardware is a Topton 4x i226v/2x 82599 with Pentium Gold 8505. When transferring from/to the Windows host, the data transfer obviously caps at 113MB/s (tested ftp). The physical interfaces (2x 2,5Gbps from the Topton, 1Gbps from Windows) are link up and advertised/linked up at 2,5Gbps on the same switch (CRS310).
Is there some known thing with the current community OPNsense version? (hope so).
Any test ideas/advice? My next thing would be to place a generic host with !=Windows and try transfer rate with because the network chip on the game rig is a ,,Killer" (RTL8125-based i think) with its own optimization software.
Thankyou!
How did you test? Speedtest, iperf? Multiple streams or one? If you really used FTP, it can normally use only one stream and is heavily limited by BDP and buffer size.
i tested so far with ftp, one stream only. Many thanks for your insight, i will continue to test and concentrate on Bandwidth-Delay Product, after i tested if this is not accidentally a pure "Killer-Feature". :) I will revert.
BTW: This was a follow-up to https://forum.opnsense.org/index.php?topic=49839.msg253185#msg253185 (i had a shaper active). But i have too much to do and cannot remember the iperf3 test results so i will have to redo them.
Remembers me on a case when one of our customer was complaining they have bad network performance. Just to find out they were using remote dd between linux systems. In the end issue was on their devices, specifically disk performance.
Not saying this is your case as well but...
For network performance always test with iperf3. Where OPN/the GW is just transport not the receiver or the sender. Do not use tests that are based and involve other components of the systems mainly disks.
Regards,
S.