[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [dm-devel] missing /sys/block/dm-*/queue



Hi Paul,

I have the same problem, I need to tune my disks to better
performance. What protocol are you using?
If you have it solve, please let me know.

Thanks

Nasc
NetApp - Enjoy it!

On Fri, May 14, 2010 at 1:43 PM, Paul Raines <raines nmr mgh harvard edu> wrote:
>
> Sorry if this is the wrong contact to ask this
>
> I am running RHEL 5.5 w/ device-mapper-multipath-0.4.7-34.el5
>
> I am trying to tune my disks for better performance with the settings
>
>  echo 2048 > /sys/block/$bdev/queue/read_ahead_kb
>  echo 64 > /sys/block/$bdev/queue/nr_requests
>  echo deadline > /sys/block/$bdev/queue/scheduler
>  echo 8192 > /sys/block/$bdev/queue/max_sectors_kb
>
> On the individual disk when I do this I get a big jump in read performance
> of nearly 2x
>
> # dd if=/dev/sdig bs=1024k of=/dev/null skip=3000 count=3000
> 3000+0 records in
> 3000+0 records out
> 3145728000 bytes (3.1 GB) copied, 13.8672 seconds, 227 MB/s
>
> # dd if=/dev/sdig bs=1024k of=/dev/null skip=9000 count=3000
> 3000+0 records in
> 3000+0 records out
> 3145728000 bytes (3.1 GB) copied, 7.70806 seconds, 408 MB/s
>
> So /dev/sdig is part of a multipath
>
> # ls -l /sys/block/dm-119/slaves/
> total 0
> lrwxrwxrwx 1 root root 0 May 14 12:14 sdez -> ../../../block/sdez
> lrwxrwxrwx 1 root root 0 May 14 12:14 sdif -> ../../../block/sdif
>
> I apply the tuning to both sdez and sdif, however, when I do the
> dd read test on the dm device, it is still slow
>
> # dd if=/dev/dm-119 bs=1024k of=/dev/null skip=18000 count=3000
> 3000+0 records in
> 3000+0 records out
> 3145728000 bytes (3.1 GB) copied, 15.1546 seconds, 208 MB/s
>
> Fine, I need to apply the tuning to the dm-119 device itself.  But there
> is no queue sub dir in /sys/block/dm-119
>
> # ls /sys/block/dm-119
> dev  holders  range  removable  size  slaves  stat  subsystem  uevent
>
>
> So how do I tune the multipath device itself?
>
> --
> ---------------------------------------------------------------
> Paul Raines                email: raines at nmr.mgh.harvard.edu
> MGH/MIT/HMS Athinoula A. Martinos Center for Biomedical Imaging
> 149 (2301) 13th Street     Charlestown, MA 02129            USA
>
>
> --
> dm-devel mailing list
> dm-devel redhat com
> https://www.redhat.com/mailman/listinfo/dm-devel
>


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]