On Fri, 2012-03-23 at 13:53 -0400, Jeff Wasilko wrote:
> On Fri, Mar 23, 2012 at 06:01:35PM +0100, Conrad Wood wrote:
> > on storage:
> > dd if=/dev/zero of=volume bs=1M : ~ 1,600MByte/s
> > cp file1 file2 : ~ 300MByte/s (both files on same volume)
> > 
> > on server
> > dd if=/dev/zero of=volume bs=1M : ~ 1,100MByte/s
> > cp file1 file2 : ~ 30MByte/s (both files on same volume)
> > 
> > Happy with the dd, but it always drops to a tenth of the performance
> > when I do it with cp.
> 
> cp doesn't use large blocks so you'll always get better performance with
> dd. Try using dd with bs=4k or 32k and see how it stacks up.
> 
> The coreutils that RHEL ships is pretty old and still uses 4K I/O. We
> upgraded to coreutils-8.13 which uses 32K blocks and it helps with
> performance. 
> 

dd with 4k still shows 680MByte/s. tbh I would expect the scheduler to
merge it together in sequential writes anyways.
I didn't find any pattern that exhibits this slow speed apart from
copying the file within the same volume.
In fact copying a file from ram on to this volume is 700+ Mbyte/s too.
I always unmount the volume after copying, just to make sure it really
is the diskspeed I am looking at not the server cache. I include the
time it takes to unmount when I calculate the speed.

Conrad


_______________________________________________
Tech mailing list
[email protected]
https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech
This list provided by the League of Professional System Administrators
 http://lopsa.org/

Reply via email to