measure the performance of fec

[zooko]

Figure out how long it takes to erasure code files of various sizes from various kinds of disk with various kinds of CPU.

[zooko]

The purpose is to compare authentication by Merkle Tree root of shares of encoded ciphertext vs. authentication by hash of ciphertext.

[zooko]

I'm doing this now (on my Mac Book Pro).

[zooko]

just read the file
N: 1048576, time: ave:    0.05,   2th-best:    0.02,   2th-worst:    0.02 (of      5), reps/s:     18, averate: 19634785
N: 4194304, time: ave:    0.06,   2th-best:    0.05,   2th-worst:    0.05 (of      5), reps/s:     17, averate: 75194678N: 16777216, time: ave:    0.29,   2th-best:    0.28,   2th-worst:    0.29 (of      5), reps/s:      3, averate: 58704285
N: 67108864, time: ave:    0.44,   2th-best:    0.29,   2th-worst:    0.54 (of      5), reps/s:      2, averate: 152414826

read the file and erasure code it
N: 1048576, time: ave:    0.13,   2th-best:    0.13,   2th-worst:    0.13 (of      5), reps/s:      7, averate:  7917270N: 4194304, time: ave:    0.45,   2th-best:    0.45,   2th-worst:    0.45 (of      5), reps/s:      2, averate:  9309219
N: 16777216, time: ave:    1.74,   2th-best:    1.74,   2th-worst:    1.74 (of      5), reps/s:      0, averate:  9664285
N: 67108864, time: ave:    7.12,   2th-best:    6.94,   2th-worst:    7.30 (of      4), reps/s:      0, averate:  9422081

read the file and hash it
N: 1048576, time: ave:    0.02,   2th-best:    0.02,   2th-worst:    0.02 (of      5), reps/s:     64, averate: 67678080
N: 4194304, time: ave:    0.05,   2th-best:    0.05,   2th-worst:    0.05 (of      5), reps/s:     18, averate: 79464075
N: 16777216, time: ave:    0.29,   2th-best:    0.29,   2th-worst:    0.30 (of      5), reps/s:      3, averate: 56985117
N: 67108864, time: ave:    0.39,   2th-best:    0.29,   2th-worst:    0.54 (of      5), reps/s:      2, averate: 172282142

read the file and erasure code it and hash the shares
N: 1048576, time: ave:    0.16,   2th-best:    0.16,   2th-worst:    0.16 (of      5), reps/s:      6, averate:  6428383
N: 4194304, time: ave:    0.58,   2th-best:    0.57,   2th-worst:    0.59 (of      5), reps/s:      1, averate:  7250374
N: 16777216, time: ave:    2.19,   2th-best:    2.19,   2th-worst:    2.19 (of      5), reps/s:      0, averate:  7675755
N: 67108864, time: ave:    8.87,   2th-best:    8.75,   2th-worst:    8.75 (of      3), reps/s:      0, averate:  7568872

[zooko]

One more time, with better formatting. The "bytes" column is the size of the file in bytes. The "secs ave:" column is the number of seconds it took to process the file. The "2nd-best" and "2nd-worst" columns are the times for 2nd-fastest and 2nd-slowest runs. The bytes/sec column is just the bytes column divided by the secs ave column.

   bytes: secs ave: 2nd-best: 2nd-worst: bytes/sec:

just read the file
 1048576      0.05      0.02       0.02   19634785
 4194304      0.06      0.05       0.05   75194678
16777216      0.29      0.28       0.29   58704285
67108864      0.44      0.29       0.54  152414826

read the file and erasure code it
 1048576      0.13      0.13       0.13    7917270
 4194304      0.45      0.45       0.45    9309219
16777216      1.74      1.74       1.74    9664285
67108864      7.12      6.94       7.30    9422081

read the file and hash it
 1048576      0.02      0.02       0.02   67678080
 4194304      0.05      0.05       0.05   79464075
16777216      0.29      0.29       0.30   56985117
67108864      0.39      0.29       0.54  172282142

read the file and erasure code it and hash the shares
 1048576      0.16      0.16       0.16    6428383
 4194304      0.58      0.57       0.59    7250374
16777216      2.19      2.19       2.19    7675755
67108864      8.87      8.75       8.75    7568872

[zooko]

I just noticed that these benchmarks show that is is faster, for sufficiently large files, to read them and hash them than just to read them. ;-)

[zooko]

<function _encode_file_not_really at 0x1365430>
N:  131072, time: ave:    0.00,   5th-best:    0.00,   5th-worst:    0.00 (of    128), reps/s:    683, averate: 89605414
N:  524288, time: ave:    0.01,   5th-best:    0.00,   5th-worst:    0.01 (of    128), reps/s:    134, averate: 70466146
N: 2097152, time: ave:    0.03,   5th-best:    0.03,   5th-worst:    0.03 (of    128), reps/s:     36, averate: 75936268
N: 8388608, time: ave:    0.12,   4th-best:    0.11,   4th-worst:    0.16 (of     45), reps/s:      8, averate: 71992492
<function _encode_file_not_really_and_hash at 0x1365470>
N:  131072, time: ave:    0.00,   5th-best:    0.00,   5th-worst:    0.00 (of    128), reps/s:    318, averate: 41695895
N:  524288, time: ave:    0.01,   5th-best:    0.01,   5th-worst:    0.01 (of    128), reps/s:     80, averate: 42066021
N: 2097152, time: ave:    0.04,   5th-best:    0.03,   5th-worst:    0.04 (of    128), reps/s:     28, averate: 59333196
N: 8388608, time: ave:    0.12,   4th-best:    0.11,   4th-worst:    0.13 (of     44), reps/s:      8, averate: 68054133
<function _encode_file at 0x1365370>
N:  131072, time: ave:    0.02,   5th-best:    0.02,   5th-worst:    0.03 (of    128), reps/s:     42, averate:  5587400
N:  524288, time: ave:    0.07,   5th-best:    0.06,   5th-worst:    0.07 (of    128), reps/s:     13, averate:  7245829
N: 2097152, time: ave:    0.24,   5th-best:    0.23,   5th-worst:    0.24 (of    112), reps/s:      4, averate:  8786003
N: 8388608, time: ave:    0.87,   4th-best:    0.87,   4th-worst:    0.87 (of     30), reps/s:      1, averate:  9633914
<function _encode_file_and_hash at 0x13654b0>
N:  131072, time: ave:    0.03,   5th-best:    0.03,   5th-worst:    0.03 (of    128), reps/s:     34, averate:  4538526
N:  524288, time: ave:    0.09,   5th-best:    0.08,   5th-worst:    0.09 (of    128), reps/s:     11, averate:  5918480
N: 2097152, time: ave:    0.30,   5th-best:    0.29,   5th-worst:    0.30 (of    102), reps/s:      3, averate:  7070671
N: 8388608, time: ave:    1.10,   4th-best:    1.09,   4th-worst:    1.10 (of     27), reps/s:      0, averate:  7651940