Skip to content

Instantly share code, notes, and snippets.

@jsmestad

jsmestad/iodata.mkd

Forked from paul/iodata.mkd
Created August 31, 2010 18:06
Show Gist options
  • Save jsmestad/559451 to your computer and use it in GitHub Desktop.
Save jsmestad/559451 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
iodata.mkd

m1.large

Single mounted EBS, XFS

Iozone: Performance Test of File I/O
        Version $Revision: 3.308 $
  Compiled for 64 bit mode.
  Build: linux

Run began: Sun Aug 22 19:35:33 2010

Include close in write timing
Include fsync in write timing
File size set to 8388608 KB
Record Size 64 KB
Command line used: iozone -c -e -s 8g -r 64k
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
       8388608      64   46115   47994    83947    82814   30989   34233   47252  3406533    47624    40921    49934   82359    85779
       8388608      64   40061   46175    88932    86837   31582   34402   44545  3422394    37983    43023    47589   86621    85250
       8388608      64   46113   51169    81848    86331   27134   31636   38385  3284538    29377    45938    48996   86642    87063

mdadm mirror

                                                          random  random    bkwd   record   stride
            KB  reclen   write rewrite    read    reread    read   write    read  rewrite     read   fwrite frewrite   fread  freread
       8388608      64   43473   46352    57438    61201   13442   29384   20835  3380154    47482    47159    47103   72094    90447
       8388608      64   46783   47332    77358    73444   31079   33088   27528  3443182    45859    42297    42109   76346    74095

mdadm stripe (2 devices)

                                                          random  random    bkwd   record   stride
            KB  reclen   write rewrite    read    reread    read   write    read  rewrite     read   fwrite frewrite   fread  freread
       8388608      64   76932   77181    99495   102703   21534   32128   35372  2647323    34882    65459    67277  102148   103247
       8388608      64   68858   69152    98342   101037   20798   29836   34225  2628765    33711    65179    68203   98138   102379
       8388608      64   61484   68382   100247   101609   20795   30242   34034  2618773    33623    65151    68490  101534   102903

mdadm raid 5 (3 devices)

Note: preferred method, for reliability and performance

                                                          random  random    bkwd   record   stride
            KB  reclen   write rewrite    read    reread    read   write    read  rewrite     read   fwrite frewrite   fread  freread
       8388608      64   42579   46492   127275   127350  121414   25241  162474  2447693   161820    47290    48201  135685   133456
       8388608      64   49093   48535   130889   133104  124248   25597  158476  2465324   163904    48986    47977  131383   130418
       8388608      64   47369   47562   130772   132806  121905   25073  155017  2070006   158502    48063    49234  128923   127998

mdadm 10

xxx

lvm stripe

Note: lvm RAID seems to be slower and less mature than mdadm

                                                          random  random    bkwd   record   stride
            KB  reclen   write rewrite    read    reread    read   write    read  rewrite     read   fwrite frewrite   fread  freread
       8388608      64   77797   89480    51364    52639   69422   59349  168993  3394045    91694    77322    86873   70346    68081
       8388608      64   82538   85382    58208    54578   71846   60353  166688  3323417    82185    85303    88210   75711    77076

lvm span

xxx

lvm stripe over mdadm mirror

Note: lvm stripes are SLOW at writes random random bkwd record stride KB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread 8388608 64 14303 30501 75643 82499 13681 13389 39655 2619943 21012 25252 36915 93384 96516

lvm span over mdadm mirror

xxx

lvm span over mdadm raid 10 (4 devices for each /dev/md)

Note: preferred method, for reliability and expandability. Poor performance for reads, however.

                                                          random  random    bkwd   record   stride
            KB  reclen   write rewrite    read    reread    read   write    read  rewrite     read   fwrite frewrite   fread  freread
       8388608      64   54380   54573    31345    32003   20892   51608  106626  3334149    31571    54076    54733   33780    31499

m2.4xlarge

Single mounted EBS, XFS

Note: The file size here was probably way to small, all the reads were just out of the buffer cache. I doubt we really got 5Gbytes/s out of EBS.

Iozone: Performance Test of File I/O
        Version $Revision: 3.308 $
  Compiled for 64 bit mode.
  Build: linux

Run began: Sun Aug 22 20:48:46 2010

Include close in write timing
Include fsync in write timing
File size set to 8388608 KB
Record Size 64 KB
Command line used: iozone -c -e -s 8g -r 64k
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
       8388608      64   55821   57627  5089526  5102975 4570300   57741 4855740  6772790  4714128    56966    57433 5076458  5077382
       8388608      64   58630   59119  5060545  5091654 4598299   55922 4822469  6830658  4746746    54273    58441 5083893  5091469
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment