Nice mod - but I think perhaps too little airflow. The mini-Kaze barely move any air at all. I believe there are other 40mm fans that might be a better compromise, yielding a bit more noise while still being a lot quieter than the original fans.
The 20mm-thick version Mini Kaze Ultra shifts a bit more air at a still-respectable 19dBA. Evercool and Sunon to 40mm PWM fans which can shift a lot more air at full whack (45+dBA) but you're at the mercy of the mainboard to PWM them down and I suspect it'll lean more towards high-noise than low. Perhaps dialing down a thicker (20, 28mm) 2/3-pin fan with a suitable controller to a compromise noise/thermal level might be a safer bet!
FYI on the fans. They are on a tray you don't have to open up the whole thing. just undo the two screws on the back and pull the module. If you are going to run PWM and less than three fans make sure you at lest plug in the one closest to the power supply. If the the fan closest to the power supply is unplugged the other two go full speed but the other two can be unplugged and it will still vary the speed appropriately.
Trying to figure out LACP. I followed the FastPath manual to add ports to the port-channels but Server 2012 still reports LACP negotiation fault. Is there something I'm missing? I haven't set subnet associations on any VLANs or IP addresses on the NIC in Windows as usually I would think LACP is a protocol negotiation that can be done without addressing.
The manual says after you add ports to LAGs you can add the LAG to the default management VLAN. If I do 'show vlan 1' it shows all the LAG groups as already being members.
Following this thread...
Been thinking of pairing one of these to my C7000 ( Using the Virtual Connect 1/10F on the network side of things ), and my Firebox ( x1250e ) is about to receive an HP NC510F. All in all 5 ports would be populated right off the bat !
UPDATE: Well, I was able to resolve my performance issues for now, but I'm not sure why this worked: I had to enable jumbo frames end-to-end to finally break through to 9Gb/s. This leads me to believe something else is off config-wise, since it seems everyone else is only seeing a 10% performance gain by using jumbo.
Anyone else seeing lower-than-expected speeds? I have a Mellanox MNPH-29D and a Chelsio T520-LL-CR connected with active DACs and can only get about 5Gb/s with 10 parallel streams in iperf. Same config on my old Brocade 8000 was hitting 9Gb/s+...
Not sure what's up. I'm going to try swapping the two DACs with two others I have, but they're all Brocade "branded" Fiberstore cables, so not sure if those might be the issue.