On Sat, 08 Jan 2005 19:33:58 -0700, Josh Coates <[EMAIL PROTECTED]> wrote:
> 
> >Pipe the output of dd into gzip and then onto the disk.
> 
> actually, this won't speed it up at all - if it has any effect, it will slow
> it down. :-(
> a disk to disk copy is dominated by the unbuffered sequential IO of the read
> and write disk (whichever takes longer - usually writes take slight longer
> than reads - but not by much - remember, this is *unbuffered* - your ondisk
> write cache will be blown, and so it's useless.)
> 
> by compressing data before you write it, you make the write side take less
> time since you are writing less data to it, but you are always still waiting
> for the read side, and you additionally chew up the CPU with compression.
> large read + write IO already chews up a significant amount of CPU.  so,
> this is not a good idea IMHO.

Except that we're talking about copying an entire mostly-empty disk. 
Writes are almost always slower than reads if they are different at
all and a great deal of space can be compressed into nearly no space
if it's all zeros.

> >Make huge files filled with zeroed blocks (dd if=/dev/zero
> of=/mnt/disk/file) and then delete them (this cleans out unused space).
> 
> this might do something useful, but i really dont understand what you are
> trying to do here. :-/

That's why you didn't understand my previous advice.

> >Adjust the block size dd uses.
> 
> this is the only thing that will make a HUGE difference.  if i remember
> correctly, the last time i did this anything over 64K (1024*64) was pretty
> good, and above that gave diminishing returns. YMMV.

Good to know.  Thanks!

--------------------
BYU Unix Users Group 
http://uug.byu.edu/ 

The opinions expressed in this message are the responsibility of their
author.  They are not endorsed by BYU, the BYU CS Department or BYU-UUG. 
___________________________________________________________________
List Info: http://uug.byu.edu/cgi-bin/mailman/listinfo/uug-list

Reply via email to