4 fio is a tool that will spawn a number of threads or processes doing a
5 particular type of io action as specified by the user. fio takes a
6 number of global parameters, each inherited by the thread unless
7 otherwise parameters given to them overriding that setting is given.
8 The typical use of fio is to write a job file matching the io load
15 fio resides in a git repo, the canonical place is:
17 git://git.kernel.dk/fio.git
19 The http protocol also works, path is the same.
21 Snapshots are frequently generated and they include the git meta data as
22 well. You can download them here:
24 http://brick.kernel.dk/snaps/
26 Pascal Bleser <guru@unixtech.be> has fio RPMs in his repository for
27 SUSE variants, you can find them here:
29 http://linux01.gwdg.de/~pbleser/rpm-navigation.php?cat=System/fio
31 Dag Wieƫrs has RPMs for Red Hat related distros, find them here:
33 http://dag.wieers.com/rpm/packages/fio/
35 Mandriva has integrated fio into their package repository, so installing
36 on that distro should be as easy as typing 'urpmi fio'.
42 There's a mailing list associated with fio. It's meant for general
43 discussion, bug reporting, questions - basically anything that has to
44 do with fio. An automated mail detailing recent commits is automatically
45 sent to the list at most daily. The list address is fio-devel@kernel.dk,
46 subscribe by sending an empty email to fio-devel+subscribe@kernel.dk.
52 Just type 'make' and 'make install'. If on FreeBSD, for now you have to
53 specify the FreeBSD Makefile with -f, eg:
55 $ make -f Makefile.Freebsd && make -f Makefile.FreeBSD install
57 Likewise with OpenSolaris, use the Makefile.solaris to compile there.
58 This might change in the future if I opt for an autoconf type setup.
65 --output Write output to file
66 --runtime Runtime in seconds
67 --latency-log Generate per-job latency logs
68 --bandwidth-log Generate per-job bandwidth logs
69 --minimal Minimal (terse) output
70 --version Print version info and exit
71 --help Print this page
72 --cmdhelp=cmd Print command help, "all" for all of them
73 --showcmd Turn a job file into command line options
74 --readonly Turn on safety read-only checks
75 --eta=when When ETA estimate should be printed
76 May be "always", "never" or "auto"
79 Any parameters following the options will be assumed to be job files,
80 unless they match a job file parameter. You can add as many as you want,
81 each job file will be regarded as a separate group and fio will stonewall
84 The --readonly switch is an extra safety guard to prevent accidentically
85 turning on a write setting when that is not desired. Fio will only write
86 if rw=write/randwrite/rw/randrw is given, but this extra safety net can
87 be used as an extra precaution. It will also enable a write check in the
88 io engine core to prevent an accidental write due to a fio bug.
94 See the HOWTO file for a more detailed description of parameters and what
95 they mean. This file contains the terse version. You can describe big and
96 complex setups with the command line, but generally it's a lot easier to
97 just write a simple job file to describe the workload. The job file format
98 is in the ini style format, as that is easy to read and write for the user.
100 The job file parameters are:
102 name=x Use 'x' as the identifier for this job.
103 description=x 'x' is a text description of the job.
104 directory=x Use 'x' as the top level directory for storing files
105 filename=x Force the use of 'x' as the filename for all files
106 in this thread. If not given, fio will make up
107 a suitable filename based on the thread and file
109 rw=x 'x' may be: read, randread, write, randwrite,
110 rw (read-write mix), randrw (read-write random mix)
111 rwmixcycle=x Base cycle for switching between read and write
113 rwmixread=x 'x' percentage of rw mix ios will be reads. If
114 rwmixwrite is also given, the last of the two will
115 be used if they don't add up to 100%.
116 rwmixwrite=x 'x' percentage of rw mix ios will be writes. See
118 rand_repeatable=x The sequence of random io blocks can be repeatable
119 across runs, if 'x' is 1.
120 size=x Set file size to x bytes (x string can include k/m/g)
121 ioengine=x 'x' may be: aio/libaio/linuxaio for Linux aio,
122 posixaio for POSIX aio, sync for regular read/write io,
123 psync for regular pread/pwrite io, mmap for mmap'ed io,
124 syslet-rw for syslet driven read/write, splice for using
125 splice/vmsplice, sgio for direct SG_IO io, net for
126 network io, or cpuio for a cycler burner load. sgio only
127 works on Linux on SCSI (or SCSI-like devices, such as
128 usb-storage or sata/libata driven) devices. Fio also has
129 a null io engine, which is mainly used for testing fio
131 iodepth=x For async io, allow 'x' ios in flight
132 overwrite=x If 'x', layout a write file first.
133 nrfiles=x Spread io load over 'x' number of files per job,
135 prio=x Run io at prio X, 0-7 is the kernel allowed range
136 prioclass=x Run io at prio class X
137 bs=x Use 'x' for thread blocksize. May include k/m postfix.
138 bsrange=x-y Mix thread block sizes randomly between x and y. May
139 also include k/m postfix.
140 direct=x 1 for direct IO, 0 for buffered IO
141 thinktime=x "Think" x usec after each io
142 rate=x Throttle rate to x KiB/sec
143 ratemin=x Quit if rate of x KiB/sec can't be met
144 ratecycle=x ratemin averaged over x msecs
145 cpumask=x Only allow job to run on CPUs defined by mask.
146 cpus_allowed=x Like 'cpumask', but allow text setting of CPU affinity.
147 fsync=x If writing with buffered IO, fsync after every
148 'x' blocks have been written.
149 end_fsync=x If 'x', run fsync() after end-of-job.
150 startdelay=x Start this thread x seconds after startup
151 runtime=x Terminate x seconds after startup. Can include a
152 normal time suffix if not given in seconds, such as
153 'm' for minutes, 'h' for hours, and 'd' for days.
154 offset=x Start io at offset x (x string can include k/m/g)
155 invalidate=x Invalidate page cache for file prior to doing io
156 sync=x Use sync writes if x and writing buffered IO.
157 mem=x If x == malloc, use malloc for buffers. If x == shm,
158 use shared memory for buffers. If x == mmap, use
160 exitall When one thread quits, terminate the others
161 bwavgtime=x Average bandwidth stats over an x msec window.
162 create_serialize=x If 'x', serialize file creation.
163 create_fsync=x If 'x', run fsync() after file creation.
164 unlink If set, unlink files when done.
165 loops=x Run the job 'x' number of times.
166 verify=x If 'x' == md5, use md5 for verifies. If 'x' == crc32,
167 use crc32 for verifies. md5 is 'safer', but crc32 is
168 a lot faster. Only makes sense for writing to a file.
169 stonewall Wait for preceeding jobs to end before running.
170 numjobs=x Create 'x' similar entries for this job
171 thread Use pthreads instead of forked jobs
173 zoneskip=y Zone options must be paired. If given, the job
174 will skip y bytes for every x read/written. This
175 can be used to gauge hard drive speed over the entire
176 platter, without reading everything. Both x/y can
177 include k/m/g suffix.
178 iolog=x Open and read io pattern from file 'x'. The file must
179 contain one io action per line in the following format:
181 where with rw=0/1 for read/write, and the offset
182 and length entries being in bytes.
183 write_iolog=x Write an iolog to file 'x' in the same format as iolog.
184 The iolog options are exclusive, if both given the
185 read iolog will be performed.
186 write_bw_log Write a bandwidth log.
187 write_lat_log Write a latency log.
188 lockmem=x Lock down x amount of memory on the machine, to
189 simulate a machine with less memory available. x can
190 include k/m/g suffix.
191 nice=x Run job at given nice value.
192 exec_prerun=x Run 'x' before job io is begun.
193 exec_postrun=x Run 'x' after job io has finished.
194 ioscheduler=x Use ioscheduler 'x' for this job.
195 cpuload=x For a CPU io thread, percentage of CPU time to attempt
197 cpuchunks=x Split burn cycles into pieces of x usecs.
203 Fio was written by Jens Axboe <axboe@kernel.dk> to enable flexible testing
204 of the Linux IO subsystem and schedulers. He got tired of writing
205 specific test applications to simulate a given workload, and found that
206 the existing io benchmark/test tools out there weren't flexible enough
207 to do what he wanted.
209 Jens Axboe <axboe@kernel.dk> 20060905