Cheatfio Disk I/O benchmark
Light · Postcard
102 × 152 mm · PDF
Dark · Postcard
102 × 152 mm · PDF
Ink saver · Postcard
102 × 152 mm · PDF
Light · A4 landscape
297 × 210 mm · PDF
Dark · A4 landscape
297 × 210 mm · PDF
Ink saver · A4 landscape
297 × 210 mm · PDF
Light · A3 landscape
420 × 297 mm · PDF
Dark · A3 landscape
420 × 297 mm · PDF
Ink saver · A3 landscape
420 × 297 mm · PDF
Start here
fio --name=t --filename=<file> --size=1g
--rw=randread --bs=4k
Random read test
fio <jobfile>
Run a job file
--ioengine=libaio --direct=1
Bypass page cache
--iodepth=<n>
Queue depth
--runtime=<s> --time_based
Run for fixed time
--numjobs=<n> --group_reporting
Parallel workers
--output-format=json
Machine output
fio --showcmd <jobfile>
Job file as CLI
Concepts
- Job file / CLI workload
- Jobs numjobs clones
- ioengine iodepth in flight
- Report IOPS, BW, latency
Running
fio <jobfile>
Run a job file
fio --name=<job> --rw=<mode> ...
Define a job on the CLI
fio --section=<name> <jobfile>
Run only one section
fio --output=<file> --output-format=<fmt>
Write report; normal/terse/json/json+
fio --minimal <jobfile>
Terse semicolon-separated output
fio --showcmd <jobfile>
Convert job file to CLI
fio --parse-only <jobfile>
Check options, no I/O
fio --readonly ...
Refuse write jobs
A safety net when pointing at real devices.
fio --cmdhelp=<option>
Help for one option
fio --enghelpList ioengines
fio --eta-newline=<time>
One progress line per interval
fio --status-interval=<time>
Print stats periodically
fio --versionShow version
Access pattern
--rw=read / write
Sequential read / write
--rw=randread / randwrite
Random read / write
--rw=randrw --rwmixread=<pct>
Mixed random I/O
--rw=rwMixed sequential I/O
--bs=<size>
Block size, e.g. 4k or 1m
--bs=<r>,<w>
Separate read,write sizes
--bsrange=<lo>-<hi>
Random block size range
--bssplit=<size>/<pct>:...
Weighted block sizes
--percentage_random=<pct>
Share of random I/O
--norandommapAllow repeated blocks
--randseed=<n>
Repeatable random sequence
--rw=trimDiscard blocks
Size and files
--filename=<path>
File or device to test
--directory=<dir>
Place generated files here
--size=<size>
I/O region per job
--io_size=<size>
Total I/O to do
--offset=<size>
Start offset in target
--nrfiles=<n>
Files per job
--filesize=<lo>-<hi>
Random file sizes
--loops=<n>
Repeat the job
--unlink=1Delete files when done
Engine and queue
--ioengine=libaio
Linux async I/O
--ioengine=io_uring
Modern Linux async I/O
--ioengine=psync
Plain pread/pwrite (default)
--ioengine=syncBasic read/write
--ioengine=posixaio
POSIX async I/O
--ioengine=nullNo I/O, test fio itself
--iodepth=<n>
I/Os kept in flight
Has little effect on synchronous engines.
--iodepth_batch_submit=<n>
Submit in batches
--direct=1O_DIRECT, skip page cache
--buffered=0Same as direct=1
--sync=1Open with O_SYNC
--fsync=<n>
fsync every N writes
--end_fsync=1fsync at job end
--invalidate=0Keep cached pages
--hipriPolled completions
--fixedbufs --registerfiles
io_uring fast path
Time and load
--runtime=<time>
Stop after this long
--time_basedLoop until runtime ends
--ramp_time=<time>
Warm-up excluded from stats
--numjobs=<n>
Parallel clones of the job
--group_reporting
One summary for all clones
--rate=<bw>
Cap bandwidth, e.g. 10m
--rate_iops=<n>
Cap IOPS
--thinktime=<us>
Pause after each I/O
--startdelay=<time>
Delay job start
--cpus_allowed=<list>
Pin to CPUs, e.g. 0-3
--latency_target=<time>
Find max load under latency
Job files
[global]ioengine=libaiodirect=1runtime=60time_based[randread-4k]rw=randreadbs=4kiodepth=32[global]Defaults for all jobs
stonewallWait for earlier jobs
--wait_for=<job>
Start after named job
exitallEnd all when one finishes
${VAR}Environment variable in job file
filename=<dev>:<dev>
Several targets
Output and logs
--output-format=json
JSON report
--output-format=normal,json
Several formats at once
--write_lat_log=<prefix>
Per-I/O latency log
--write_bw_log=<prefix>
Bandwidth log
--write_iops_log=<prefix>
IOPS log
--log_avg_msec=<ms>
Average log entries
--lat_percentiles=1
Report total latency percentiles
--verify=crc32c --do_verify=1
Check data read back
--refill_buffers
New data for every write
--client=<host> / --server
Remote run
Recipes
fio --name=randread --filename=<file> --size=2g \ --rw=randread --bs=4k --ioengine=libaio --iodepth=32 \ --direct=1 --runtime=60 --time_based --group_reportingfio --name=seqwrite --filename=<file> --size=4g \ --rw=write --bs=1m --ioengine=libaio --iodepth=8 \ --direct=1 --end_fsync=1fio --name=mixed --filename=<file> --size=2g \ --rw=randrw --rwmixread=70 --bs=8k --iodepth=16 \ --direct=1 --numjobs=4 --runtime=60 --time_based \ --group_reporting --output-format=json --output=mixed.jsonfio --name=lat --filename=<file> --size=1g \ --rw=randread --bs=4k --iodepth=1 --direct=1 \ --runtime=30 --time_based --lat_percentiles=1
