fileio tests show output in bits but numbers are byte values #12

leonroy opened this Issue Oct 29, 2015 · 3 comments


None yet

2 participants

leonroy commented Oct 29, 2015

For example running the following over a Gibabit link:

sysbench --test=fileio --file-total-size=4G --file-test-mode=seqwr --max-time=240 --max-requests=0 --file-block-size=1M --num-threads=4 --file-fsync-all run


sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 4

Extra file open flags: 0
128 files, 32Mb each
4Gb total file size
Block size 1Mb
Calling fsync() after each write operation.
Using synchronous I/O mode
Doing sequential write (creation) test
Threads started!

Operations performed:  0 Read, 4096 Write, 4096 Other = 8192 Total
Read 0b  Written 4Gb  Total transferred 4Gb  (91.604Mb/sec)
   91.60 Requests/sec executed

Test execution summary:
    total time:                          44.7140s
    total number of events:              4096
    total time taken by event execution: 175.7213
    per-request statistics:
         min:                                 25.53ms
         avg:                                 42.90ms
         max:                                174.82ms
         approx.  95 percentile:              51.47ms

Threads fairness:
    events (avg/stddev):           1024.0000/0.71
    execution time (avg/stddev):   43.9303/0.00

Note that the file sizes are in Mb, Gb, Kb, whereas the numerical values are the actual MB value. Is this a bug in sysbench? Has it been fixed?


Yes, it's a known issue which has not been fixed yet. Generally, sysbench is neither consistent, nor standards compliant with respect to reporting size units. Sometimes Kb/Mb/Gb are used as units, and sometimes KB/MB/GB. In both cases they actually mean KiB/MiB/GiB as defined by IEC:

leonroy commented Oct 29, 2015

Thanks for the info. And really appreciate the tool, it's awesome. Just wrote a post on how to use it:

@akopytov akopytov added the bug label Mar 18, 2016

Fixed in 1.0 with #32.

@akopytov akopytov closed this Mar 23, 2016
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment