On 12/16/2009 9:34 AM, Chan Chung Hang Christopher wrote:
Steve Thompson wrote:
On Wed, 16 Dec 2009, Chan Chung Hang Christopher wrote:
Steve Thompson wrote:
On Tue, 15 Dec 2009, Scott Ehrlich wrote:
I have a client with a handful of USB drives connected to a CentOS box. I am charged with binding the USB drives together into a single LVM for a cheap storage data pool (10 x 1 TB usb drives = 10 TB cheap storage in a single mount point).
I tried doing this for fun once upon a time, using 6 1TB drives. I can save you a lot of grief by suggesting that you don't think about this any further. Boy is it slow. And extremely unreliable. And slow. Don't even do it for backups. Did I say it was slow?
Please qualify 'slow'. Was it dog slow, turtle-slow, snail-slow or slowaris slow?
Slower than all of those. Top write speed I could ever achieve with a USB-2 interface and SATA drives was 20 MB/sec with a trailing wind, and usually half of that, with a single stream. I even tried USB-1 for more laughs; 1 MB/sec on a truly good day. With multiple writers, performance dropped so far as to be unusable (below 1 MB/sec). And we're talking mkfs times in _days_. The host was a CentOS 5.2 box, 32-bit.
Kudos to Steve for proving that USB2's 480mbits/sec is really just a sham.
Now I wonder if you can daisy chain IEEE1394 devices...or try out eSATA...:-P _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Any host based technology won't get you half of the claimed speed with any kind of reliability. I don't think ti will ever really outrun something like SATAII or SAS. What makes it funnier is Intel is saying this will make external RAID on USB possible...just keep in mind FRIAD and that's what USB RAID really is.