====== sio ntap von NetApp ======

Dieser Benchmark ist von NetApp und arbeitet ähnlich wie dd

Hier ein paar globale Infos dazu:
[[https://communities.netapp.com/blogs/zurich/2011/01/24/sio--performance-load-testing-simplified]]

Download:
[[http://now.netapp.com/NOW/download/tools/sio_ntap/]]

 
**Install unter Linux:**
<code>
mkdir -p sio_ntap
tar -xvzf sio_ntap.tar.gz -C sio_ntap
make linux
</code>

----

**Syntax:**
<code>
Version: 3.00
Usage:
  sio <read%> <rand%> <blksz> <file_size> <seconds> <numthreads> <dev> [devs...]

<read>:       percentage of accesses that are reads.  Range [0,100].
              BEWARE, writing to a file is unchecked and will trash files
<rand>:       percentage of acceses that are random.  Range [0,100].
              Sequential accesses = 0% random
<blksz>:      size of I/O's. Example: 2k, 4k, 1m
<file_size>:  total bytes accessed in each file (e.g. 100m, 2g, 1000k)
<seconds>:    Runtime for test.  Counting starts AFTER all threads have started
<numthreads>: Concurrent I/O generators.  Uses real individual threads.
<dev>:        Device to access.  May be file (foo.out) or device (/dev/dsk/etc)
[devs...]:    Multiple devices can be specified.  I/O is distributed evenly and
              randomly across the devices.

Options:
  -noflock           :  prevents the files from being locked during access
  -output -filename  :  send all output from the command to 'filename'
  -direct            :  disable file system caching - available in aix, solaris and linux

Examples:
  Random 4K Reads to files a.file, b.file for 10 seconds with 2 threads.
  Accessing a total of 100 megabytes in each file.  File locks off.
        sio_ntap 100 100 4096 100m 10 2 a.file b.file -noflock
  Or by saving the output to file 'foo.out'
        sio_ntap 100 100 4096 100m 10 2 a.file b.file -noflock -output -foo.out

Notes:
  This program supports these OS's: Windows, Solaris, Linux, HPUX, AIX
</code>

----

**Benchmark-Beispiele:**

Write Traffic. No Ramdom. 8Kb Block Size. 512Mb File. 300 Sekunden lang. 50 Threads. Direct IO
<code>
./sio_ntap_linux 0 0 8k 512m 300 50 /tmp/sio_test -direct
</code>

ReadTraffic. No Ramdom. 8Kb Block Size. 512Mb File. 300 Sekunden lang. 50 Threads. Direct IO
<code>
./sio_ntap_linux 100 0 8k 512m 300 50 /tmp/sio_test -direct
</code>

----

==== Resultate von Systemen: ====
=== Backup01: HW-Raid5 mit 5x SATA 2Tb (DELL) ===
^ Blocksize ^Write   ^ Read   ^ IOPS   ^ FileSize ^ DirectIO ^
| 8k        | 140Mb   |        | 17453  | 1024Mb    | Yes |
| 8k        |         | 506Mb  | 63315  | 1024Mb    | Yes |

----

=== Troja: SSD M500 (128Gb) on SATA1 (Q6600) ===
^ Blocksize ^Write   ^ Read   ^ IOPS   ^ FileSize ^ DirectIO ^
| 8k        | 100Mb   |        | 12613  | 512Mb     | Yes |
| 8k        |         | 253Mb  | 31671  | 512Mb     | Yes |

----

=== Kirby: VM auf Bowser (HP DL380G6) unter vSphere 5.5 (4x SATA 500Gb 2,5" Raid 10) ===
^ Blocksize ^Write   ^ Read   ^ IOPS   ^ FileSize ^ DirectIO ^
| 8k        | 58Mb    |        | 7270   | 2048Mb    | Yes |
| 8k        |         | 90Mb   | 11217  | 2048Mb    | Yes |
| 8k        | 76Mb    |        | 9530   | 2048Mb    | No  |
| 8k        |         | 135Mb  | 16967  | 2048Mb    | No  |

----

=== Bender: VM auf Jarvis (Selbstbau) unter vSphere 5.5 U2 (4x SATA 3Tb 3,5" Raid 5 mit LSI 9260-i8) ===
^ Blocksize ^Write   ^ Read   ^ IOPS   ^ FileSize ^ DirectIO ^
| 8k        | 8Mb     |        | 999    | 2048Mb    | Yes |
| 256k      | 44Mb    |        | 176    | 2048Mb    | Yes | 
| 8k        |         | 473M   | 59221  | 2048Mb    | Yes |
| 8k        | 38Mb    |        | 4691   | 2048Mb    | No  |
| 256k      | 53Mb    |        | 205    | 2048Mb    | No  |