I am about to give up on the Brocade 1020 cards. Based on my iperf tests, in order to get any performance near 10G, I need to change the TCP Window size. When I do that, I am getting 4 to 6G. Running the iperf test with the default window size (64K), I am getting terrible results.
What to do next? I have spent so much time and money trying to a 10G connection from my VM Hosts to my SAN server. Infiniband did not work out for either. Need help.
As someone within the thread already mentioned, you should consider to run your iperf tests with more threads.
I'm not sure but at the screenshot you delivered I can see that you only used standard (1 thread). Isn't it? maybe I'm wrong.
I always have to take around 10,20 or more threads (the -p option for iperf) to get close to 10G but without changing TCP windows sizes at all.
I can't give you sources but I'm pretty sure changing windows sizes isn't so useful as you might think.
It's doesn't matter what kind of card/technique I was using. 10G Intel Ethernet stuff vs. Brocade.
the lasts test i run with the brocade cards between 2 Windows Server 2012 R2 i chose this to get 10G.
C:\iperf-2.0.5-2-win32>iperf.exe -c 192.168.x.x -P 30 -f MBytes -t 1000000
Let us know what happened.
Update:
You're using 2008 R2 and Windows 7?
Maybe there are limitations.
Another problem can be the hardware itself.
Maybe they're are limitations of the PCIe slot or RAM or CPU at you systems.
PCIe 2.0 with minimum 4x should be fast enough but.
The Brocades work with PCIe 2.0 and 8 lanes. Total throughput is above 10G.
Windows 7 works with more energy saving options.
Did you tried to turn everything off on hardware and Windows?
maximum power on windows and hardware speedsteps for example?
It would be interesting what you exactly use on both sides.