On Thu, Jun 2, 2011 at 5:50 PM, Jeffrey Bastian <jbastian(a)redhat.com> wrote:
On 2011-05-31 10:00, Jeffrey Bastian wrote:
> I recently ran some bonnie++ benchmarks against both an SD card and a
> USB 2.0 hard drive and I was seeing about 10 MB/sec on the USB drive.
> This was with the 2.6.35-g6d019da-dirty kernel.
> I'll have to try this again with the pings and see if that improves the
I ran bonnie++ again on my PandaBoard with a USB hard drive and I also
saw the performance double when I was pinging the system. I'm now
getting 18 MB/sec writes and 24 MB/sec reads.
I posted the full bonnie++ results at:
Just wondering if someone got on with this. Maybe a tcpdump from the
USB-traffic helps. I'd suggest to compare the following
a) tcpdump from the USB-storage without pinging
b) tcpdump from the USB-storage with pinging (or other usage)
tcpdump can capture the USB-traffic (on a similar level as Ethernet),
with a command like this:
# tcpdump -w /tmp/my-usb-traffic.pcap -s 256 -i usbmon0
usbmon0 captures all USB-traffic, where usbmon<N> captures the traffic
from USB-bus <N>. The USB-bus where the device is connected can be
found in /proc/bus/usb/devices.
tshark (or wireshark) can read the generated .pcap file. In order to
make sense out of the bus-numbers and device-IDs, the
/proc/bus/usb/devices is very valuable. In case someone decides to
make traces available, please provide that file as well.