[clug] fastest way to copy a fs
eyal at eyal.emu.id.au
Wed Jan 17 09:19:16 GMT 2007
Alex Satrapa wrote:
> On 17 Jan 2007, at 15:45, Richard Reynolds wrote:
>>> Having the disks on different controllers means the bottleneck is
>>> halving the bandwidth of the PCI bus rather than halving the
>>> bandwidth of the drive controller.
>> really only a bottleneck if the controller is half as fast as the PCI
>> bus. 133MB/s is a hard game to catch with sustained reads/writes
> Exactly my point - even a USB 2.0 controller will have idle time when
> it's only controlling one drive. Strictly speaking the bottleneck
> becomes the drive itself (most drives I've played with max out at
> 20MB/s off the platters). One you have more than that "off-the- platter"
> bandwidth from source to destination, you have very little room left
> for optimising transfer speed.
>> feel free to, but benchmarks are published for many of these already
>> google turned up a ton of results that seemed right on.
> Can you share the search? I've looked for "optimising file system
> transfer" and "fastest way to copy one disk to another" but they don't
> turn up anything interesting...
So far this is what I see. It is the fs that spreads the writing so that
we are nowhere near sequential. A short test copied / (about 6GB) in some
Eyal Lebedinsky (eyal at eyal.emu.id.au) <http://samba.org/eyal/>
attach .zip as .dat
More information about the linux