I do mention in my tagline that this blog may contain some information about computers, so you’ve been warned. The following article is quite technical, and may not be of interest to the photography contingent that might otherwise be interested in my blog articles. I posted this because I spent several days attempting a particular task, and managed to get it down to a few simple instructions. It may be useful for other attempting the same task.
I’ve recently set up an Infiniband fabric at home, and had a lot of trouble getting it working on Ubuntu. Windows 7 was a breeze, I just instlled the OFED drivers from openfbrics.org.
Here’s the steps to get it working on Ubuntu 10.10. It might not seem a lot, but there was a lot of messing to get it down to these few steps.
Install Ubuntu. Everything below is done as root – “sudo bash”.
Add a file to /etc/udev/rules.d called, say, 99-udev-umad.rules. This will cause the correct entries to be created in /sys, which otherwise do not get created for Ubuntu 10.10.
Insert the following (Make sure you check the cut-and-paste, some people find that the quotes get changed):
KERNEL==”umad*”, NAME=”infiniband/%k”, MODE=”0666″
KERNEL==”issm*”, NAME=”infiniband/%k”, MODE=”0666″
KERNEL==”uverbs*”, NAME=”infiniband/%k”, MODE=”0666″
Edit /etc/modules and add the following modules:
Next, “apt-get install opensm”. This will install the subnet manager and all the relevant dependencies, libibverbs, etc.
Then add the relevant entries for the interface into /etc/network/interfaces file:
iface ib0 inet static
Then reboot. This will create the relevant infiniband entries in /sys, load the ipoib modules, and bring up the infiniband port with an ip address.
You should now have a functioning infiniband port on your Ubuntu machine.
However, there is still some more investigation to be done. Initially, when I was mixing custom kernels with OFED drivers, and stock linux kernel drivers, netperf was showing 7 gbps throughput. With the configurtation above, it’s down at about 25 mbps. Dreadfully slow. I’ll have to find out what optimisations are needed (or other drivers) in order to get the speed back up to 7 gbps.
Note: iperf maxed out at 1.2 gbps, and on the current linux install, I couldnt get netperf client working at all. netserver would work, but only showed a throughput of 25mbps from a Win7 client. HOWEVER, when I set up the raid with 6 old 160G drives, the “hdparm -t /dev/md0p1″ showed 250MB/sec reads, and I got the same from the Win7 machine using samba across the infiniband fabric. This seems to indicate that iperf and netperf are completely unreliable for testing this type of connection. Bear in mind though that I did have netperf running on the previous ubuntu install, but that installation was so messy I don’t know what drivers and user-space software was running. I reckon it’s the kind of think that may be fixed in the stock Ubuntu install in the near future. For the moment, just go with real-world testing, i.e. copying large files from ramdisk to ramdisk, for example.