[Novalug] iozone benchmarking

Megan Larko larkoc@iges.org
Wed Jan 16 15:57:42 EST 2008


Megan Larko wrote:

Okay.  I got iozone3_283.tar and installed it onto two systems: 
1--larkobox  Ubuntu Gutsy Gibbon	2--oss3  CentOS 5 final (8 CPUs)
cat /proc/cpuinfo
processor       : 0                     : 7
vendor_id       : GenuineIntel          : GenuineIntel
cpu family      : 15                    : 6
model           : 1                     ; 15
model name      : Intel(R) Pentium(R) 4 CPU 1.80GHz
  stepping        : 2
                                         : Intel(R) Xeon(R) CPU
                                           E5345  @ 2.33GHz
                                         : 7 (stepping)
cpu MHz         : 1800.135              : 2327.564
cache size      : 256 KB                : 4096 KB


cat /proc/meminfo
MemTotal:       514316 kB		: 16308556 kB
MemFree:         70336 kB		: 15704804 kB
Buffers:         11672 kB		: 169832 kB
Cached:         122980 kB		: 235504 kB
SwapCached:      22760 kB		: 44 kB
Active:         311024 kB		: 248332 kB
Inactive:        41164 kB		: 220560 kB

Very different boxes!

The larkobox has only one smallish disk:
larkoc@larkoc:~/CREW/benchmarks$ df -h
Filesystem            Size  Used Avail Use% Mounted on
/dev/sda2             6.9G  4.2G  2.4G  64% /
varrun                252M   96K  252M   1% /var/run
varlock               252M     0  252M   0% /var/lock
udev                  252M   68K  252M   1% /dev
devshm                252M     0  252M   0% /dev/shm
lrm                   252M   34M  218M  14% 
/lib/modules/2.6.22-14-generic/volatile
/dev/sda1              96M   60M   31M  66% /boot
/dev/sda4              28G   14G   13G  53% /home

The oss3 box has many Hitachi 1Tb Ultrastar disks:
[larkoc@oss3 current]$ df -h
Filesystem            Size  Used Avail Use% Mounted on
/dev/sda1             4.9G  3.3G  1.4G  70% /
tmpfs                 7.8G     0  7.8G   0% /dev/shm
crew01:/home          1.4T  412G  894G  32% /home
/dev/sdc1             5.5T  187M  5.4T   1% /LSI1
/dev/sdd1             5.4T  187M  5.4T   1% /LSI2
/dev/mapper/oss3space-LV1
                       5.4T  191M  5.4T   1% /SCRATCH

I ran iozone on the command-line and seemed to get essentially the same 
results for both systems.   I'm not buying it.

larkobox OUTPUT:
         Run began: Wed Jan 16 15:31:09 2008

         Excel chart generation enabled
         Auto Mode
         Using maximum file size of 4 kilobytes.
         Command line used: ./iozone -Ra -g 1Gb -i 0 -i 1
         Output is in Kbytes/sec
         Time Resolution = 0.000001 seconds.
         Processor cache size set to 1024 Kbytes.
         Processor cache line size set to 32 bytes.
         File stride size set to 17 * record size.
                                                             random 
random    bkwd  record  stride
               KB  reclen   write rewrite    read    reread    read 
write    read rewrite    read   fwrite frewrite   fread  freread

iozone test complete.
Excel output is below:

"Writer report"
         "4"  "8"  "16"  "32"  "64"  "128"  "256"  "512"  "1024"  "2048" 
  "4096"  "8192"  "16384"
"0"   0

"Re-writer report"
         "4"  "8"  "16"  "32"  "64"  "128"  "256"  "512"  "1024"  "2048" 
  "4096"  "8192"  "16384"
"0"   0

"Reader report"
         "4"  "8"  "16"  "32"  "64"  "128"  "256"  "512"  "1024"  "2048" 
  "4096"  "8192"  "16384"
"0"   0

"Re-Reader report"
         "4"  "8"  "16"  "32"  "64"  "128"  "256"  "512"  "1024"  "2048" 
  "4096"  "8192"  "16384"

**********************************************************************
oss3 OUTPUT:
         Run began: Wed Jan 16 15:53:49 2008

         Excel chart generation enabled
         Auto Mode
         Using maximum file size of 24 kilobytes.
         Command line used: ./iozone -Ra -g 24Gb -i 0 -i 1
         Output is in Kbytes/sec
         Time Resolution = 0.000001 seconds.
         Processor cache size set to 1024 Kbytes.
         Processor cache line size set to 32 bytes.
         File stride size set to 17 * record size.
                                                             random 
random    bkwd  record  stride
               KB  reclen   write rewrite    read    reread    read 
write    read rewrite    read   fwrite frewrite   fread  freread

iozone test complete.
Excel output is below:

"Writer report"
         "4"  "8"  "16"  "32"  "64"  "128"  "256"  "512"  "1024"  "2048" 
  "4096"  "8192"  "16384"
"0"   0

"Re-writer report"
         "4"  "8"  "16"  "32"  "64"  "128"  "256"  "512"  "1024"  "2048" 
  "4096"  "8192"  "16384"
"0"   0

"Reader report"
         "4"  "8"  "16"  "32"  "64"  "128"  "256"  "512"  "1024"  "2048" 
  "4096"  "8192"  "16384"
"0"   0

What am I missing in the issuing of the iozone command?

TIA,
megan


> Hello,
> 
> I would like to get some benchmarks on my filesystem on my work 
> computer.  It is a 16-slot JBOD with an LSI edge card.  The drives in 
> the 16 bays are all sATA Hitachi Ultrastar 1Tb units.   I used parted to 
> partition them.   I used mke2fs -j -T largefile -m 1 for one group of 
> drives passed to the CentOS 5 system.   The other group of drives I did 
> the same mke2fs but with out the -T largefile option.   Using 8 drives 
> each, the size formatted was 5.4Tb and 5.5Tb respectively.  I am 
> wondering about the performance of the set-up.  How does this new LSI 
> edge card compare with the internal Sil cards in the other boxes?  Does 
> the "largefile" option make any difference?     I may even scrap the 
> 5.4Tb ext3 and try the experimental ext4 or ZFS file system. 
> I am leaning toward using bonnie for this to get some numbers to compare 
> amongst file systems on the same disks (which will hopefully have some 
> meaning).
> 
> Is there a good bonnie HOWTO out there.  My searches didn't come up with 
> much in the way of usage of bonnie, only benchmark results done by 
> others.  What would be good arguments to supply to the test?
> 
> Thanks group!
> megan
> _______________________________________________
> Novalug mailing list
> Novalug@calypso.tux.org
> http://calypso.tux.org/cgi-bin/mailman/listinfo/novalug




More information about the Novalug mailing list