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