the test values are not correct. the seeker utility used the current number of seconds as the seed for rand() and so the processes used the same random offsets. the number is 10000 seeks/second on all 16 usb drives when reading from /dev/sd*, 800 when reading from a 64GB LVM2 logical volume. And then 6000 when seeking in the files on 3 partially filled LVM2 volumes. The reason for the last 2 timings is not yet clear to me.
On Fri, Jan 15, 2010 at 11:13:33AM +0100, sascha wrote: > On Fri, Jan 15, 2010 at 05:03:03AM +0100, sascha wrote: > > > USB is widely available, but still a serial bus, so only a single > > > request to a single memory stick per host controller is in flight at > > > any one time. (Even with 8 USB ports it's still one controller.) > > > > tests show otherwise: > > single usb stick: > > Results: 967 seeks/second, 1.03 ms random access time > > > > 3 usb sticks on a hub accessed concurrently: > > Results: 907 seeks/second, 1.10 ms random access time > > Results: 916 seeks/second, 1.09 ms random access time > > Results: 915 seeks/second, 1.09 ms random access time > > > a ran a test on my 16 usb-stick linear raid, with 4 sticks per hub, > 4 hubs, 1 on the first usb port of my computer, 3 on another hub connected > to the second usb port of the computer. i can do 16000 seeks per second > and transfer 512 bytes on each seek, giving 62.5 usec combined access time > for the raid. it also blinks like a christmas tree ;) > _______________________________________________ > A51 mailing list > [email protected] > http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51 _______________________________________________ A51 mailing list [email protected] http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51
