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/
31 Starting with Debian "Squeeze", fio packages are part of the official
32 Debian repository. http://packages.debian.org/search?keywords=fio
35 Starting with Ubuntu 10.04 LTS (aka "Lucid Lynx"), fio packages are part
36 of the Ubuntu "universe" repository.
37 http://packages.ubuntu.com/search?keywords=fio
40 Pascal Bleser <guru@unixtech.be> has fio RPMs in his repository for SUSE
41 variants, you can find them here:
42 http://linux01.gwdg.de/~pbleser/rpm-navigation.php?cat=System/fio
45 Dag Wieƫrs has RPMs for Red Hat related distros, find them here:
46 http://dag.wieers.com/rpm/packages/fio/
49 Mandriva has integrated fio into their package repository, so installing
50 on that distro should be as easy as typing 'urpmi fio'.
53 Packages for Solaris are available from OpenCSW. Install their pkgutil
54 tool (http://www.opencsw.org/get-it/pkgutil/) and then install fio via
61 There's a mailing list associated with fio. It's meant for general
62 discussion, bug reporting, questions, and development - basically anything
63 that has to do with fio. An automated mail detailing recent commits is
64 automatically sent to the list at most daily. The list address is
65 fio@vger.kernel.org, subscribe by sending an email to
66 majordomo@vger.kernel.org with
70 in the body of the email. Archives can be found here:
72 http://www.spinics.net/lists/fio/
74 and archives for the old list can be found here:
76 http://maillist.kernel.dk/fio-devel/
82 Just type 'make' and 'make install'. If on FreeBSD, for now you have to
83 specify the FreeBSD Makefile with -f and use gmake (not make), eg:
85 $ gmake -f Makefile.Freebsd && gmake -f Makefile.FreeBSD install
89 $ gmake -f Makefile.aix && gmake -f Makefile.aix install
91 Likewise with OpenSolaris, use the Makefile.solaris to compile there.
92 The OpenSolaris make should work fine. This might change in the
93 future if I opt for an autoconf type setup.
95 If your compile fails with an error like this:
98 In file included from fio.h:23,
100 os/os.h:15:20: error: libaio.h: No such file or directory
101 In file included from gettime.c:8:
102 fio.h:119: error: field 'iocb' has incomplete type
103 make: *** [gettime.o] Error 1
105 Check that you have the libaio development package installed. On RPM
106 based distros, it's typically called libaio-devel.
113 --debug Enable some debugging options (see below)
114 --output Write output to file
115 --timeout Runtime in seconds
116 --latency-log Generate per-job latency logs
117 --bandwidth-log Generate per-job bandwidth logs
118 --minimal Minimal (terse) output
119 --version Print version info and exit
120 --help Print this page
121 --cmdhelp=cmd Print command help, "all" for all of them
122 --showcmd Turn a job file into command line options
123 --readonly Turn on safety read-only checks, preventing writes
124 --eta=when When ETA estimate should be printed
125 May be "always", "never" or "auto"
126 --section=name Only run specified section in job file
127 --alloc-size=kb Set smalloc pool to this size in kb (def 1024)
130 Any parameters following the options will be assumed to be job files,
131 unless they match a job file parameter. You can add as many as you want,
132 each job file will be regarded as a separate group and fio will stonewall
135 The --readonly switch is an extra safety guard to prevent accidentically
136 turning on a write setting when that is not desired. Fio will only write
137 if rw=write/randwrite/rw/randrw is given, but this extra safety net can
138 be used as an extra precaution. It will also enable a write check in the
139 io engine core to prevent an accidental write due to a fio bug.
141 The debug switch allows adding options that trigger certain logging
142 options in fio. Currently the options are:
144 process Dump info related to processes
145 file Dump info related to file actions
146 io Dump info related to IO queuing
147 mem Dump info related to memory allocations
148 blktrace Dump info related to blktrace setup
149 verify Dump info related to IO verification
150 all Enable all debug options
151 random Dump info related to random offset generation
152 parse Dump info related to option matching and parsing
153 diskutil Dump info related to disk utilization updates
154 job:x Dump info only related to job number x
155 mutex Dump info only related to mutex up/down ops
156 profile Dump info related to profile extensions
157 time Dump info related to internal time keeping
158 ? or help Show available debug options.
160 You can specify as many as you want, eg --debug=file,mem will enable
161 file and memory debugging.
163 The section switch is meant to make it easier to ship a bigger job file
164 instead of several smaller ones. Say you define a job file with light,
165 moderate, and heavy parts. Then you can ask fio to run the given part
166 only by giving it a --section=heavy command line option. The section
167 option only applies to job sections, the reserved 'global' section is
168 always parsed and taken into account.
170 Fio has an internal allocator for shared memory called smalloc. It
171 allocates shared structures from this pool. The pool defaults to 1024k
172 in size, and can grow to 128 pools. If running large jobs with randommap
173 enabled it can run out of memory, in which case the --alloc-size switch
174 is handy for starting with a larger pool size. The backing store is
175 files in /tmp. Fio cleans up after itself, while it is running you
176 may see .fio_smalloc.* files in /tmp.
182 See the HOWTO file for a more detailed description of parameters and what
183 they mean. This file contains the terse version. You can describe big and
184 complex setups with the command line, but generally it's a lot easier to
185 just write a simple job file to describe the workload. The job file format
186 is in the ini style format, as that is easy to read and write for the user.
188 The job file parameters are:
190 name=x Use 'x' as the identifier for this job.
191 description=x 'x' is a text description of the job.
192 directory=x Use 'x' as the top level directory for storing files
193 filename=x Force the use of 'x' as the filename for all files
194 in this thread. If not given, fio will make up
195 a suitable filename based on the thread and file
197 rw=x 'x' may be: read, randread, write, randwrite,
198 rw (read-write mix), randrw (read-write random mix)
199 rwmixcycle=x Base cycle for switching between read and write
201 rwmixread=x 'x' percentage of rw mix ios will be reads. If
202 rwmixwrite is also given, the last of the two will
203 be used if they don't add up to 100%.
204 rwmixwrite=x 'x' percentage of rw mix ios will be writes. See
206 rand_repeatable=x The sequence of random io blocks can be repeatable
207 across runs, if 'x' is 1.
208 size=x Set file size to x bytes (x string can include k/m/g)
209 ioengine=x 'x' may be: aio/libaio/linuxaio for Linux aio,
210 posixaio for POSIX aio, solarisaio for Solaris
211 native async IO, windowsaio for Windows native async IO,
212 sync for regular read/write io,
213 psync for regular pread/pwrite io, vsync for regular
214 readv/writev (with queuing emulation) mmap for mmap'ed
215 io, syslet-rw for syslet driven read/write, splice for
216 using splice/vmsplice, sg for direct SG_IO io, net
217 for network io, or cpuio for a cycler burner load. sg
218 only works on Linux on SCSI (or SCSI-like devices, such
219 as usb-storage or sata/libata driven) devices. Fio also
220 has a null io engine, which is mainly used for testing
223 iodepth=x For async io, allow 'x' ios in flight
224 overwrite=x If 'x', layout a write file first.
225 nrfiles=x Spread io load over 'x' number of files per job,
227 prio=x Run io at prio X, 0-7 is the kernel allowed range
228 prioclass=x Run io at prio class X
229 bs=x Use 'x' for thread blocksize. May include k/m postfix.
230 bsrange=x-y Mix thread block sizes randomly between x and y. May
231 also include k/m postfix.
232 direct=x 1 for direct IO, 0 for buffered IO
233 thinktime=x "Think" x usec after each io
234 rate=x Throttle rate to x KB/sec
235 ratemin=x Quit if rate of x KB/sec can't be met
236 ratecycle=x ratemin averaged over x msecs
237 cpumask=x Only allow job to run on CPUs defined by mask.
238 cpus_allowed=x Like 'cpumask', but allow text setting of CPU affinity.
239 fsync=x If writing with buffered IO, fsync after every
240 'x' blocks have been written.
241 end_fsync=x If 'x', run fsync() after end-of-job.
242 startdelay=x Start this thread x seconds after startup
243 runtime=x Terminate x seconds after startup. Can include a
244 normal time suffix if not given in seconds, such as
245 'm' for minutes, 'h' for hours, and 'd' for days.
246 offset=x Start io at offset x (x string can include k/m/g)
247 invalidate=x Invalidate page cache for file prior to doing io
248 sync=x Use sync writes if x and writing buffered IO.
249 mem=x If x == malloc, use malloc for buffers. If x == shm,
250 use shared memory for buffers. If x == mmap, use
252 exitall When one thread quits, terminate the others
253 bwavgtime=x Average bandwidth stats over an x msec window.
254 create_serialize=x If 'x', serialize file creation.
255 create_fsync=x If 'x', run fsync() after file creation.
256 unlink If set, unlink files when done.
257 loops=x Run the job 'x' number of times.
258 verify=x If 'x' == md5, use md5 for verifies. If 'x' == crc32,
259 use crc32 for verifies. md5 is 'safer', but crc32 is
260 a lot faster. Only makes sense for writing to a file.
261 For other types of checksumming, see HOWTO.
262 stonewall Wait for preceeding jobs to end before running.
263 numjobs=x Create 'x' similar entries for this job
264 thread Use pthreads instead of forked jobs
266 zoneskip=y Zone options must be paired. If given, the job
267 will skip y bytes for every x read/written. This
268 can be used to gauge hard drive speed over the entire
269 platter, without reading everything. Both x/y can
270 include k/m/g suffix.
271 iolog=x Open and read io pattern from file 'x'. The file must
272 contain one io action per line in the following format:
274 where with rw=0/1 for read/write, and the offset
275 and length entries being in bytes.
276 write_iolog=x Write an iolog to file 'x' in the same format as iolog.
277 The iolog options are exclusive, if both given the
278 read iolog will be performed.
279 write_bw_log Write a bandwidth log.
280 write_lat_log Write a latency log.
281 lockmem=x Lock down x amount of memory on the machine, to
282 simulate a machine with less memory available. x can
283 include k/m/g suffix.
284 nice=x Run job at given nice value.
285 exec_prerun=x Run 'x' before job io is begun.
286 exec_postrun=x Run 'x' after job io has finished.
287 ioscheduler=x Use ioscheduler 'x' for this job.
288 cpuload=x For a CPU io thread, percentage of CPU time to attempt
290 cpuchunks=x Split burn cycles into pieces of x usecs.
297 Fio works on (at least) Linux, Solaris, AIX, OSX, NetBSD, and FreeBSD. Some
298 features and/or options may only be available on some of the platforms,
299 typically because those features only apply to that platform (like the
300 solarisaio engine, or the splice engine on Linux).
302 Some features are not available on FreeBSD/Solaris even if they could be
303 implemented, I'd be happy to take patches for that. An example of that is
304 disk utility statistics and (I think) huge page support, support for that
305 does exist in FreeBSD/Solaris.
307 Fio uses pthread mutexes for signalling and locking and FreeBSD does not
308 support process shared pthread mutexes. As a result, only threads are
309 supported on FreeBSD. This could be fixed with sysv ipc locking or
310 other locking alternatives.
312 Other *BSD platforms are untested, but fio should work there almost out
313 of the box. Since I don't do test runs or even compiles on those platforms,
314 your mileage may vary. Sending me patches for other platforms is greatly
315 appreciated. There's a lot of value in having the same test/benchmark tool
316 available on all platforms.
318 Note that POSIX aio is not enabled by default on AIX. If you get messages like:
320 Symbol resolution failed for /usr/lib/libc.a(posix_aio.o) because:
321 Symbol _posix_kaio_rdwr (number 2) is not exported from dependent module /unix.
323 you need to enable POSIX aio. Run the following commands as root:
325 # lsdev -C -l posix_aio0
326 posix_aio0 Defined Posix Asynchronous I/O
327 # cfgmgr -l posix_aio0
328 # lsdev -C -l posix_aio0
329 posix_aio0 Available Posix Asynchronous I/O
331 POSIX aio should work now. To make the change permanent:
333 # chdev -l posix_aio0 -P -a autoconfig='available'
340 Fio was written by Jens Axboe <axboe@kernel.dk> to enable flexible testing
341 of the Linux IO subsystem and schedulers. He got tired of writing
342 specific test applications to simulate a given workload, and found that
343 the existing io benchmark/test tools out there weren't flexible enough
344 to do what he wanted.
346 Jens Axboe <axboe@kernel.dk> 20060905