Great tip, Alan. I'll look into capturing USB traffic with Wireshark as well. Both fio and Wireshark should be available for Linux and Windows. On Fri, Jun 17, 2016 at 10:11 PM, Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote: > On Fri, 17 Jun 2016, Michael Schaller wrote: > >> Alan, thank you for the analysis and you are correct in your >> assumption that this is a plain dumb thumb drive. The puzzling thing >> for me is that the drive only shows this bottleneck under Linux and >> not under Max OS X or Windows. Here are my results for Linux vs. Mac >> OS X again: >> >> Write speed under Linux: >> $ sudo dd if=/dev/zero of=/dev/sdb bs=1M count=256 >> 256+0 records in >> 256+0 records out >> 268435456 bytes (268 MB) copied, 5.98705 s, 44.8 MB/s >> >> Write speed under Mac OS X: >> $ sudo dd if=/dev/zero of=/dev/rdisk2 bs=1048576 count=256 >> 256+0 records in >> 256+0 records out >> 268435456 bytes transferred in 0.941998 secs (284963933 bytes/sec) > > That _is_ unexpected. It's possible that this doesn't reflect the time > required to perform the writes, but merely the time to store the data > in a memory buffer. The actual USB transfer might take place after the > dd program terminates. > >> Alan, do you have any idea why the write speed differs so heavily >> between Linux (~45 MB/s) and the other OSs (~270 MB/s)? Could the >> initialization of the flash drive be different under Linux and the >> other OSs? > > I can't tell for sure. Not without seeing something equivalent to a > usbmon trace for OS X or Windows. > > Is Wireshark on either platform is capable of capturing USB traffic? > If it is, you could run the same kind of test as before under one of > those operating systems and post the results. (And even if Wireshark > can't do this, there are publicly available programs that can.) > > Alan Stern > -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
