To answer your questions:
1. we did setup "blockdev":
# blockdev --getra /dev/sdc3
4096
2. when we using dump for 16GB data (first logical volume to third logical
volume) and it take 26 min to finish (10MB/sec).
#dump -0uf /u3/BKOS/root.dmp -B 40000000 /
DUMP: Date of this level 0 dump: Fri Dec 4 09:03:57 2009
DUMP: Dumping /dev/sdc2 (/) to /u3/BKOS/root.dmp
DUMP: Label: /
DUMP: Writing 10 Kilobyte records
DUMP: mapping (Pass I) [regular files]
DUMP: mapping (Pass II) [directories]
DUMP: estimated 15803002 blocks on 0.40 tape(s).
DUMP: Volume 1 started with block 1 at: Fri Dec 4 09:04:24 2009
DUMP: dumping (Pass III) [directories]
DUMP: dumping (Pass IV) [regular files]
DUMP: 16.69% done at 8453 kB/s, finished in 0:24
DUMP: 36.04% done at 9305 kB/s, finished in 0:17
DUMP: 57.06% done at 9887 kB/s, finished in 0:11
DUMP: 76.35% done at 9955 kB/s, finished in 0:06
DUMP: 99.68% done at 10418 kB/s, finished in 0:00
DUMP: Closing /u3/BKOS/root.dmp
DUMP: Volume 1 completed at: Fri Dec 4 09:31:23 2009
DUMP: Volume 1 16771020 blocks (16377.95MB)
DUMP: Volume 1 took 0:26:59
DUMP: Volume 1 transfer rate: 10358 kB/s
DUMP: 16771020 blocks (16377.95MB) on 1 volume(s)
DUMP: finished in 1607 seconds, throughput 10436 kBytes/sec
DUMP: Date of this level 0 dump: Fri Dec 4 09:03:57 2009
DUMP: Date this dump completed: Fri Dec 4 09:31:23 2009
DUMP: Average transfer rate: 10358 kB/s
DUMP: DUMP IS DONE
--- 09/12/3 (四)，Bond Masuda <[email protected]> 寫道：
> 寄件者: Bond Masuda <[email protected]>
> 主旨: RE: R900 internal disks I/O slow!!!!
> 收件者: "'mcclnx mcc'" <[email protected]>, [email protected]
> 日期: 2009年12月3日,四,下午8:33
> Ok. That is very slow.
>
> Keep in mind, since you have 2 mirror sets, and 1 single
> disk, your configuration will not be faster than any single
> disk. So, typical SAS drive can maybe do about 80MB/sec, so
> you should expect somewhere between 60-100MB/sec i think..
> but that is a guess.
>
> How are you measuring the throughput when you copy a file?
> Is it something like:
>
> rsync -av /path/to/source/file /path/to/destination/file
>
> Can you do a read-only or write-only test instead? Maybe
> using dd if=/dev/zero of=/path/to/logical/volume/testfile
> bs=1024 count=10240
>
> When you do your copy of file, what is the output of
> 'iostat' ? can you run 'iostat -m 1 10' during the test?
>
> Also, one thing that I've notice makes a big performance
> difference on the SAS/SATA controllers in Dell servers is
> the block device read-ahead cache I/O scheduler. To see what
> your read-ahead cache looks like:
>
> blockdev --getra /dev/sda (or whatever the device is)
>
> To set it:
>
> blockdev --setra 8192 /dev/sda (set it to 8MB)
>
> I don't know about mirror sets, but with RAID-5 sets,
> read-ahead tuning made a big difference in throughput
> tests.
>
> Also, try changing to the 'deadline' I/O scheduler on the
> device and see if that helps.
>
> What is the file system configuration? ext3 or?
>
> > -----Original Message-----
> > From: mcclnx mcc [mailto:[email protected]]
> > Sent: Thursday, December 03, 2009 5:11 PM
> > To: [email protected];
> Bond Masuda
> > Subject: RE: R900 internal disks I/O slow!!!!
> >
> > sorry it is typo. it is "8 to 9 MB/sec"
> >
> >
> > --- 09/12/3 (四)，Bond Masuda <[email protected]>
> 寫道：
> >
> > > 寄件者: Bond Masuda <[email protected]>
> > > 主旨: RE: R900 internal disks I/O slow!!!!
> > > 收件者: "'mcclnx mcc'" <[email protected]>,
> [email protected]
> > > 日期: 2009年12月3日,四,下午4:54
> > > Hi,
> > >
> > > 8 to 9GB/sec is slow?
> > >
> > > -Bond
> > >
> > > > -----Original Message-----
> > > > From: [email protected]
> > > [mailto:linux-poweredge-
> > > > [email protected]]
> > > On Behalf Of mcclnx mcc
> > > > Sent: Thursday, December 03, 2009 1:51 PM
> > > > To: [email protected]
> > > > Subject: R900 internal disks I/O slow!!!!
> > > >
> > > > we have several R900 server with Redhat 5.3
> O.S. in
> > > it. R900 have 5
> > > > internal 450GB SAS disks. we
> configured it as 0
> > > and 1 mirror, 2 and 3
> > > > mirror, last disk singe.
> > > >
> > > > I have been perform file in between each
> logical disks
> > > and I/O is very
> > > > slow (something like 8 to 9GB/sec).
> Even my OLD
> > > DELL 2650 servers
> > > > (Redhat 4.7) internal disks run fast than
> it.
> > > >
> > > > Does anyone know why?
> > > >
> > > > Thanks.
> > > >
> > > >
> ___________________________________________________
> > > > 蠟腔汜魂撈媆籵 ㄜ
> > >
> 𩆨籵﹜𠸉猸﹜汜魂﹜馱釬珨棒詻隅ㄐ
> > > > http://messenger.yahoo.com.tw/
> > > >
> > > >
> _______________________________________________
> > > > Linux-PowerEdge mailing list
> > > > [email protected]
> > > > https://lists.us.dell.com/mailman/listinfo/linux-poweredge
> > > > Please read the FAQ at http://lists.us.dell.com/faq
> > >
> > >
> >
> > ___________________________________________________
> > 您的生活即時通 －
> 溝通、娛樂、生活、工作一次搞定！
> > http://messenger.yahoo.com.tw/
>
>
___________________________________________________
您的生活即時通 － 溝通、娛樂、生活、工作一次搞定！
http://messenger.yahoo.com.tw/
_______________________________________________
Linux-PowerEdge mailing list
[email protected]
https://lists.us.dell.com/mailman/listinfo/linux-poweredge
Please read the FAQ at http://lists.us.dell.com/faq
