rkoberman at gmail.com
Fri Jan 17 21:17:52 UTC 2014
On Fri, Jan 17, 2014 at 8:51 AM, paul beard <paulbeard at gmail.com> wrote:
> Grr. The return key should not be the same as Send.
> I had some problems with clonehdd that turned out to be hardware issued: my
> disks has somehow defaulted to PIO4 from UDMA which made this process run
> so slow it would just quit with finishing.
> Now things are more what I expect, completing a clone of my root disk in
> 524 minutes, down from multiple days…
> But it still seems slow if the disks are 1.5Gb/sec SATA disks. Looking
> around, it looks like this might be useful to speed things up.
> Paul Beard / www.paulbeard.org/
Remember that 1.5GB/s is the speed supported by the electronics and the
cache. It is not the speed that the disk actually reads or writes from/to
the platters. When cloning, very little of the data is in cache, so you are
generally limited by seek times (should be minimal if the code is well
done) and rotational speed. The really then boils down to transfer speeds
are going ot be close to what is possible with the rotational speed.
Since I lack specifics on your disk drive, you'll have to do the
arithmetic. It will only be approximate because of the tricks disks use for
spatial optimization, but it won't be 1.5GB/s. Also be aware that disabling
write cache will significantly reduce performance. (It should be "on" by
N.B. Several statements above are incomplete or slightly inaccurate. I
think they are close enough for this discussion and being both complete and
accurate would make this message way too long.
R. Kevin Oberman, Network Engineer, Retired
E-mail: rkoberman at gmail.com
More information about the freebsd-ports