xfs
[Top] [All Lists]

Re: xfs cpu usage

To: Greg Freemyer <freemyer@xxxxxxxxxxxxxxxxx>
Subject: Re: xfs cpu usage
From: Michal Adamczak <pokryfka@xxxxxxxxxxxxxxxxxx>
Date: Wed, 19 Mar 2003 19:16:50 +0100
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <20030319175011.RXTR25382.imf58bis.bellsouth.net@tiger2>
References: <20030319175011.RXTR25382.imf58bis.bellsouth.net@tiger2>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mutt/1.5.3i
thanks for prompt answer

k, i know scsi devices are generally more independent and as such less
cpu consuming

but this drive is on my desktop and i really apprectiate buying 80GB for
$100

on a production host (where there are both scsi and ide drives for less
critical information) i still use ext3

i thought about switching to xfs and wanted to make some
preliminary tests on my desktop 

unfortunately the way it works right now is unacceptable

On Wed, Mar 19, 2003 at 01:00:44PM -0500, Greg Freemyer wrote:
> I thought this was the very reason people used SCSI in preference to IDE.
> 
> i.e. IDE uses the CPU to control seeks, SCSI just tells the controller, and 
> the controller interrupts the CPU when the seek is done.
> 
> Obviously you could upgrade to SCSI.
> 
> Alternatively, the 3ware controllers present the CPU with a SCSI interface, 
> but talk to IDE drives on the back-end.
> 
> I assume there are other controllers that do this as well.  
> 
> I know that some of the Promise controllers imply they do this, but in 
> reality they don't reduce CPU load.
> 
> Other Promise controllers may do the job right.
> 
> FYI: A 2-port 3ware card (IIRC 7000-2) only costs a little over $100 (US) and 
> they are supported in the vanilla kernel.
> 
> Greg
>  >>  hi,
>  >>  when coping/deleting files (both small and very large)
>  >>  the cpu usage reaches 100% so that it's hardly possible to do anything
>  >>  else at the same time
> 
>  >>  what could possibly be the reason?
> 
>  >>  i use linux 2.5.65 (though same goes for 2.4.20)
>  >>  debian sid
> 
>  >>  i use the following hdparm settings:
>  >>  #hdparm -d1 c1 -u1 -m8 /dev/hda
> 
>  >>  which results in 
> 
>  >>  #hdparm  /dev/hda
>  >>  
>  >>  /dev/hda:
>  >>  multcount    = 16 (on)
>  >>      IO_support   =  1 (32-bit)
>  >>      unmaskirq    =  1 (on)
>  >>      using_dma    =  1 (on)
>  >>      keepsettings =  0 (off)
>  >>      readonly     =  0 (off)
>  >>      readahead    = 256 (on)
>  >>      geometry     = 23989/16/63, sectors = 156301488, start = 0
>  >>                   
> 
>  >>  -- 
>  >>  Michal Adamczak
>  >>  pokryfka @ druid.if.uj.edu.pl
>  >>  GS dpu C+++ UL+++++ L+++(++++) w--- !tv h*
>  >>  I code AND I bathe. (seen on slashdot)

-- 
Michal Adamczak
pokryfka @ druid.if.uj.edu.pl
GS dpu C+++ UL+++++ L+++(++++) w--- !tv h*
I code AND I bathe. (seen on slashdot)


<Prev in Thread] Current Thread [Next in Thread>