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, and development - basically anything
44 that has to do with fio. An automated mail detailing recent commits is
45 automatically sent to the list at most daily. The list address is
46 fio@vger.kernel.org, subscribe by sending an email to
47 majordomo@vger.kernel.org with
51 in the body of the email. Archives can be found here:
53 http://www.spinics.net/lists/fio/
55 and archives for the old list can be found here:
57 http://maillist.kernel.dk/fio-devel/
63 Just type 'make' and 'make install'. If on FreeBSD, for now you have to
64 specify the FreeBSD Makefile with -f and use gmake (not make), eg:
66 $ gmake -f Makefile.Freebsd && gmake -f Makefile.FreeBSD install
68 Likewise with OpenSolaris, use the Makefile.solaris to compile there.
69 The OpenSolaris make should work fine. This might change in the
70 future if I opt for an autoconf type setup.
72 If your compile fails with an error like this:
75 In file included from fio.h:23,
77 os/os.h:15:20: error: libaio.h: No such file or directory
78 In file included from gettime.c:8:
79 fio.h:119: error: field 'iocb' has incomplete type
80 make: *** [gettime.o] Error 1
82 Check that you have the libaio development package installed. On RPM
83 based distros, it's typically called libaio-devel.
90 --debug Enable some debugging options (see below)
91 --output Write output to file
92 --timeout Runtime in seconds
93 --latency-log Generate per-job latency logs
94 --bandwidth-log Generate per-job bandwidth logs
95 --minimal Minimal (terse) output
96 --version Print version info and exit
97 --help Print this page
98 --cmdhelp=cmd Print command help, "all" for all of them
99 --showcmd Turn a job file into command line options
100 --readonly Turn on safety read-only checks, preventing writes
101 --eta=when When ETA estimate should be printed
102 May be "always", "never" or "auto"
103 --section=name Only run specified section in job file
104 --alloc-size=kb Set smalloc pool to this size in kb (def 1024)
107 Any parameters following the options will be assumed to be job files,
108 unless they match a job file parameter. You can add as many as you want,
109 each job file will be regarded as a separate group and fio will stonewall
112 The --readonly switch is an extra safety guard to prevent accidentically
113 turning on a write setting when that is not desired. Fio will only write
114 if rw=write/randwrite/rw/randrw is given, but this extra safety net can
115 be used as an extra precaution. It will also enable a write check in the
116 io engine core to prevent an accidental write due to a fio bug.
118 The debug switch allows adding options that trigger certain logging
119 options in fio. Currently the options are:
121 process Dump info related to processes
122 file Dump info related to file actions
123 io Dump info related to IO queuing
124 mem Dump info related to memory allocations
125 blktrace Dump info related to blktrace setup
126 verify Dump info related to IO verification
127 all Enable all debug options
128 random Dump info related to random offset generation
129 parse Dump info related to option matching and parsing
130 diskutil Dump info related to disk utilization updates
131 job:x Dump info only related to job number x
132 mutex Dump info only related to mutex up/down ops
133 profile Dump info related to profile extensions
134 time Dump info related to internal time keeping
135 ? or help Show available debug options.
137 You can specify as many as you want, eg --debug=file,mem will enable
138 file and memory debugging.
140 The section switch is meant to make it easier to ship a bigger job file
141 instead of several smaller ones. Say you define a job file with light,
142 moderate, and heavy parts. Then you can ask fio to run the given part
143 only by giving it a --section=heavy command line option. The section
144 option only applies to job sections, the reserved 'global' section is
145 always parsed and taken into account.
147 Fio has an internal allocator for shared memory called smalloc. It
148 allocates shared structures from this pool. The pool defaults to 1024k
149 in size, and can grow to 128 pools. If running large jobs with randommap
150 enabled it can run out of memory, in which case the --alloc-size switch
151 is handy for starting with a larger pool size. The backing store is
152 files in /tmp. Fio cleans up after itself, while it is running you
153 may see .fio_smalloc.* files in /tmp.
159 See the HOWTO file for a more detailed description of parameters and what
160 they mean. This file contains the terse version. You can describe big and
161 complex setups with the command line, but generally it's a lot easier to
162 just write a simple job file to describe the workload. The job file format
163 is in the ini style format, as that is easy to read and write for the user.
165 The job file parameters are:
167 name=x Use 'x' as the identifier for this job.
168 description=x 'x' is a text description of the job.
169 directory=x Use 'x' as the top level directory for storing files
170 filename=x Force the use of 'x' as the filename for all files
171 in this thread. If not given, fio will make up
172 a suitable filename based on the thread and file
174 rw=x 'x' may be: read, randread, write, randwrite,
175 rw (read-write mix), randrw (read-write random mix)
176 rwmixcycle=x Base cycle for switching between read and write
178 rwmixread=x 'x' percentage of rw mix ios will be reads. If
179 rwmixwrite is also given, the last of the two will
180 be used if they don't add up to 100%.
181 rwmixwrite=x 'x' percentage of rw mix ios will be writes. See
183 rand_repeatable=x The sequence of random io blocks can be repeatable
184 across runs, if 'x' is 1.
185 size=x Set file size to x bytes (x string can include k/m/g)
186 ioengine=x 'x' may be: aio/libaio/linuxaio for Linux aio,
187 posixaio for POSIX aio, solarisaio for Solaris
188 native async IO, sync for regular read/write io,
189 psync for regular pread/pwrite io, vsync for regular
190 readv/writev (with queuing emulation) mmap for mmap'ed
191 io, syslet-rw for syslet driven read/write, splice for
192 using splice/vmsplice, sg for direct SG_IO io, net
193 for network io, or cpuio for a cycler burner load. sg
194 only works on Linux on SCSI (or SCSI-like devices, such
195 as usb-storage or sata/libata driven) devices. Fio also
196 has a null io engine, which is mainly used for testing
199 iodepth=x For async io, allow 'x' ios in flight
200 overwrite=x If 'x', layout a write file first.
201 nrfiles=x Spread io load over 'x' number of files per job,
203 prio=x Run io at prio X, 0-7 is the kernel allowed range
204 prioclass=x Run io at prio class X
205 bs=x Use 'x' for thread blocksize. May include k/m postfix.
206 bsrange=x-y Mix thread block sizes randomly between x and y. May
207 also include k/m postfix.
208 direct=x 1 for direct IO, 0 for buffered IO
209 thinktime=x "Think" x usec after each io
210 rate=x Throttle rate to x KB/sec
211 ratemin=x Quit if rate of x KB/sec can't be met
212 ratecycle=x ratemin averaged over x msecs
213 cpumask=x Only allow job to run on CPUs defined by mask.
214 cpus_allowed=x Like 'cpumask', but allow text setting of CPU affinity.
215 fsync=x If writing with buffered IO, fsync after every
216 'x' blocks have been written.
217 end_fsync=x If 'x', run fsync() after end-of-job.
218 startdelay=x Start this thread x seconds after startup
219 runtime=x Terminate x seconds after startup. Can include a
220 normal time suffix if not given in seconds, such as
221 'm' for minutes, 'h' for hours, and 'd' for days.
222 offset=x Start io at offset x (x string can include k/m/g)
223 invalidate=x Invalidate page cache for file prior to doing io
224 sync=x Use sync writes if x and writing buffered IO.
225 mem=x If x == malloc, use malloc for buffers. If x == shm,
226 use shared memory for buffers. If x == mmap, use
228 exitall When one thread quits, terminate the others
229 bwavgtime=x Average bandwidth stats over an x msec window.
230 create_serialize=x If 'x', serialize file creation.
231 create_fsync=x If 'x', run fsync() after file creation.
232 unlink If set, unlink files when done.
233 loops=x Run the job 'x' number of times.
234 verify=x If 'x' == md5, use md5 for verifies. If 'x' == crc32,
235 use crc32 for verifies. md5 is 'safer', but crc32 is
236 a lot faster. Only makes sense for writing to a file.
237 For other types of checksumming, see HOWTO.
238 stonewall Wait for preceeding jobs to end before running.
239 numjobs=x Create 'x' similar entries for this job
240 thread Use pthreads instead of forked jobs
242 zoneskip=y Zone options must be paired. If given, the job
243 will skip y bytes for every x read/written. This
244 can be used to gauge hard drive speed over the entire
245 platter, without reading everything. Both x/y can
246 include k/m/g suffix.
247 iolog=x Open and read io pattern from file 'x'. The file must
248 contain one io action per line in the following format:
250 where with rw=0/1 for read/write, and the offset
251 and length entries being in bytes.
252 write_iolog=x Write an iolog to file 'x' in the same format as iolog.
253 The iolog options are exclusive, if both given the
254 read iolog will be performed.
255 write_bw_log Write a bandwidth log.
256 write_lat_log Write a latency log.
257 lockmem=x Lock down x amount of memory on the machine, to
258 simulate a machine with less memory available. x can
259 include k/m/g suffix.
260 nice=x Run job at given nice value.
261 exec_prerun=x Run 'x' before job io is begun.
262 exec_postrun=x Run 'x' after job io has finished.
263 ioscheduler=x Use ioscheduler 'x' for this job.
264 cpuload=x For a CPU io thread, percentage of CPU time to attempt
266 cpuchunks=x Split burn cycles into pieces of x usecs.
273 Fio works on (at least) Linux, Solaris, and FreeBSD. Some features and/or
274 options may only be available on some of the platforms, typically because
275 those features only apply to that platform (like the solarisaio engine, or
276 the splice engine on Linux).
278 Some features are not available on FreeBSD/Solaris even if they could be
279 implemented, I'd be happy to take patches for that. An example of that is
280 disk utility statistics and (I think) huge page support, support for that
281 does exist in FreeBSD/Solaris.
283 Fio uses pthread mutexes for signalling and locking and FreeBSD does not
284 support process shared pthread mutexes. As a result, only threads are
285 supported on FreeBSD. This could be fixed with sysv ipc locking or
286 other locking alternatives.
288 Other *BSD platforms are untested, but fio should work there almost out
289 of the box. Since I don't do test runs or even compiles on those platforms,
290 your mileage may vary. Sending me patches for other platforms is greatly
291 appreciated. There's a lot of value in having the same test/benchmark tool
292 available on all platforms.
299 Fio was written by Jens Axboe <axboe@kernel.dk> to enable flexible testing
300 of the Linux IO subsystem and schedulers. He got tired of writing
301 specific test applications to simulate a given workload, and found that
302 the existing io benchmark/test tools out there weren't flexible enough
303 to do what he wanted.
305 Jens Axboe <axboe@kernel.dk> 20060905