Re: Performance degradation when using partitions

[Bill Davidsen <davidsen@tmr.com>]

Daniel Nilsson wrote:

> On Tue, Nov 22, 2005 at 10:48:41AM -0500, Bill Davidsen wrote:
>  
>
> > Daniel Nilsson wrote:
> >    
> >
> > > While setting up a software RAID-5 array I started looking into the
> > > performance aspect of using partioned drives versus the whole disks
> > > for a RAID-5 array. I have an Adaptec 2400a controller which through
> > > the I2O kernel driver gives me access to 4x 250GB disks (JBOD mode).
> > >      
> > Did you get an answer on this? And does it happen if you use the drives 
> > directly, /dev/hdN or /dev/sdN instead of using I2O? I didn't see an 
> > obvious speed penalty in raw access of drives vs. partitions, but I 
> > lacked the hardware to really match your setup, particularly the I2O use 
> > vs. direct access to /dev/sd[ef].
> >    
>
> Bill,
>
> No, I didn't get an answer on this. I've done some more experiments
> with the drives, but since they are connected to an Adaptec 2400A RAID
> controller (in JBOD mode) I need to go through some I2O driver in or
> to see the drives at all. So I never have direct access to these
> drives as /dev/hdN or /dev/sdN. There are however two different
> drivers available for this RAID controller, one is the standard I2O
> driver and the other one is the Adaptec dpt_i2o driver.
>
> The results are the same though whether I use the Linux I2O driver or
> the Adaptec dpt_i2o, the software raid array is rebuilding at roughly
> half the speed when the drives are partioned. I don't know what other
> data to provide in ordet to get any further in the testing.
>  

Unfortunately, until someone else has the time and interest to make more 
tests I don't see a good next step. I might be able to at least do some 
better quantified tests later this weekend, but I have no I2O devices, 
so it isn't going to tell much.

--
bill davidsen <davidsen@tmr.com>
 CTO TMR Associates, Inc
 Doing interesting things with small computers since 1979

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/