login
Header Space

 
 

Re: DELL PERC 5iR slow performance

Previous thread: calling syscalls directly from asm by Vincent GROSS on Saturday, July 14, 2007 - 12:48 pm. (3 messages)

Next thread: explanation of pf state-policy floating vs if-bound? by Bohdan Tashchuk on Saturday, July 14, 2007 - 7:25 pm. (1 message)
To: <misc@...>
Date: Saturday, July 14, 2007 - 1:32 pm

Hi,

I have a DELL SC440 box with a DELL SAS 5i/R Raid controller.
Installing OpenBSD 4.1 on the box, everything looks good except for
the DELL SAS controller performance.

Have tried to search the mailling list archive as well as build the
latest current kernel but doesnt help.   Attached is the dmesg and
bonie++ test on the RAID 1 virtual disk.


-- 
./Brandon
Version  1.03       ------Sequential Output------ --Sequential Input- --Random-
                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
puffy.bsdway.n 512M  6893   3  7070   0  6306   1 59161  47 59821   5 144.9   0
                    ------Sequential Create------ --------Random Create--------
                    -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                 16    63   0 +++++ +++   118   0    59   0 +++++ +++   152   0
puffy.bsdway.net,512M,6893,3,7070,0,6306,1,59161,47,59821,5,144.9,0,16,63,0,+++++,+++,118,0,59,0,+++++,+++,152,0



OpenBSD 4.1 (GENERIC.MP) #1225: Sat Mar 10 19:23:18 MST 2007
    deraadt@i386.openbsd.org:/usr/src/sys/arch/i386/compile/GENERIC.MP
cpu0: Intel(R) Xeon(R) CPU 3050 @ 2.13GHz ("GenuineIntel" 686-class) 2.13 GHz
cpu0: FPU,V86,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CFLUSH,DS,ACPI,MMX,FXSR,SSE,SSE2,SS,HTT,TM,SBF,
SSE3,MWAIT,DS-CPL,VMX,EST,TM2,CX16,xTPR
real mem  = 1071718400 (1046600K)
avail mem = 970432512 (947688K)
using 4278 buffers containing 53710848 bytes (52452K) of memory
mainbus0 (root)
bios0 at mainbus0: AT/286+ BIOS, date 03/28/07, BIOS32 rev. 0 @ 0xffe90, SMBIOS rev. 2.3 @ 0xf0450 (63 entries)
bios0: Dell Inc. PowerEdge SC440
apm0 at bios0: Power Management spec V1.2
apm0: AC on, battery charge unknown
apm0: flags 30102 dobusy 0 doidle 1
pcibios0 at bios0: rev 2.1 @ 0xf0000/0x10000
pcibios0: PCI IRQ Routing Table rev 1.0 @ 0xfed10/256 (14 ent...
To: Brandon Lee <brandon@...>
Cc: <misc@...>
Date: Saturday, July 14, 2007 - 2:19 pm

I bet write cache is disabled.  I don't recall if you can enable it in
the bios but have a look.
To: Marco Peereboom <slash@...>
Cc: <misc@...>
Date: Saturday, July 14, 2007 - 3:32 pm

Hi Marco,

There is no option to enable write cache via the PERC 5i/R controller
or Bios.  From bios, i could only select the perfornance of the HDD as
in "Bypass, Quiet, Suggested and Performance".

-- 
./Brandon
To: Brandon Lee <brandon@...>
Cc: Marco Peereboom <slash@...>, <misc@...>
Date: Sunday, July 15, 2007 - 2:38 am

I've seen this before. For some stupid reason you cannot enable the
write cache from the BIOS. What a collegue did is use a linux live CD
with a management program from LSI to enable it. After that all was
OK.
To: <misc@...>
Date: Sunday, July 15, 2007 - 10:17 am

I use LSI FiberChannel cards at work and have seen similar issues where
a parameter can't be changed via the BIOS interface.  I ended up making a
FreeDOS boot CD so I could use the LSI management tool for things that
weren't available from the BIOS.

g.day
To: Brandon Lee <brandon@...>
Cc: Marco Peereboom <slash@...>, <misc@...>
Date: Sunday, July 15, 2007 - 3:03 am

BTW, I found the instuctions from my collegue here. 
Quick translation form dutch:

Get linux drivers from dell site: mptlinux-3.02.83.12-5dkms.noarch.rpm
Put the drivers on a USB stick.
Boot knoppix live cd (http://ftp.knoppix.nl/os/Linux/distr/knoppix/KNOPPIX_V5.1.1CD-2007-01-04-EN.iso)

Then in knoppix:

sudo -s
mount /dev/sda1 (the USB stick)
cd /media/sda1
cp mptlinux-3.02.83.12-5dkms.noarch.rpm  /root
cd /root
alien -i mptlinux-3.02.83.12-5dkms.noarch.rpm
cd /ramdisk/usr/src/mtplinux-3.02.83/lsi
tar xvfz lsiutil.tgz
cd lsiutil
make
./lsiutil

choose controller (er is er maar een)
choose RAID functions 
choose change Volume settings
first option: Write Cache Enabled (Yes/No default No): Yes
remaining thing default
close and reboot.

ugly but it works.

	-Otto
To: Otto Moerbeek <otto@...>
Cc: Marco Peereboom <slash@...>, <misc@...>
Date: Sunday, July 15, 2007 - 1:44 pm

Hi Otto,

Here is the bonnie++ test result after i enabled the write cache.
Seems good but not too sure if that is the best performance.

any comments?  anyway the PERC 5i/R Raid Card is doing RAID 1 with 2
SATA II drives.

-- 
./Brandon



Version  1.03       ------Sequential Output------ --Sequential Input- --Random-
                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
puffy            2G 44136  24 48902  17 23327   7 51902  39 50587   7 132.4   0
                    ------Sequential Create------ --------Random Create--------
                    -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                 16  1901  41 +++++ +++ +++++ +++  2519  56  6131 100 12760  99
puffy           ,2G,44136,24,48902,17,23327,7,51902,39,50587,7,132.4,0,16,1901,41,+++++,+++,+++++,+++,2519,56,6131,100,12760,99
To: Brandon Lee <brandon@...>
Cc: Otto Moerbeek <otto@...>, <misc@...>
Date: Sunday, July 15, 2007 - 10:12 pm

You pay for sata you get sata; no magic there.
To: Otto Moerbeek <otto@...>
Cc: Brandon Lee <brandon@...>, <misc@...>
Date: Sunday, July 15, 2007 - 10:08 am

i'll look at this to see if we can add it to the driver.
To: <misc@...>
Date: Sunday, July 15, 2007 - 10:19 am

that would be cool
To: Otto Moerbeek <otto@...>
Cc: Marco Peereboom <slash@...>, <misc@...>
Date: Sunday, July 15, 2007 - 6:43 am

Hi Otto,

Really thanks alot.  Will try it few hours later as im out.  Will
revert if this works for me! :)

-- 
./Brandon
Previous thread: calling syscalls directly from asm by Vincent GROSS on Saturday, July 14, 2007 - 12:48 pm. (3 messages)

Next thread: explanation of pf state-policy floating vs if-bound? by Bohdan Tashchuk on Saturday, July 14, 2007 - 7:25 pm. (1 message)
speck-geostationary