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 If you are inside a corporate firewall, git:// may not always work for
20 you. In that case you can use the http protocol, path is the same:
22 http://git.kernel.dk/fio.git
24 Snapshots are frequently generated and they include the git meta data as
25 well. You can download them here:
27 http://brick.kernel.dk/snaps/
34 Starting with Debian "Squeeze", fio packages are part of the official
35 Debian repository. http://packages.debian.org/search?keywords=fio
38 Starting with Ubuntu 10.04 LTS (aka "Lucid Lynx"), fio packages are part
39 of the Ubuntu "universe" repository.
40 http://packages.ubuntu.com/search?keywords=fio
43 Pascal Bleser <guru@unixtech.be> has fio RPMs in his repository for SUSE
44 variants, you can find them here:
45 http://linux01.gwdg.de/~pbleser/rpm-navigation.php?cat=System/fio
48 Dag Wieƫrs has RPMs for Red Hat related distros, find them here:
49 http://dag.wieers.com/rpm/packages/fio/
52 Mandriva has integrated fio into their package repository, so installing
53 on that distro should be as easy as typing 'urpmi fio'.
56 Packages for Solaris are available from OpenCSW. Install their pkgutil
57 tool (http://www.opencsw.org/get-it/pkgutil/) and then install fio via
61 Bruce Cran <bruce@cran.org.uk> has fio packages for Windows at
62 http://www.bluestop.org/fio .
68 There's a mailing list associated with fio. It's meant for general
69 discussion, bug reporting, questions, and development - basically anything
70 that has to do with fio. An automated mail detailing recent commits is
71 automatically sent to the list at most daily. The list address is
72 fio@vger.kernel.org, subscribe by sending an email to
73 majordomo@vger.kernel.org with
77 in the body of the email. Archives can be found here:
79 http://www.spinics.net/lists/fio/
81 and archives for the old list can be found here:
83 http://maillist.kernel.dk/fio-devel/
89 Just type 'make' and 'make install'.
91 Note that GNU make is required. On BSD it's available from devel/gmake;
92 on Solaris it's in the SUNWgmake package. On platforms where GNU make
93 isn't the default, type 'gmake' instead of 'make'.
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.
112 On Windows Cygwin (http://www.cygwin.com/) is required in order to
113 build fio. To create an MSI installer package install WiX 3.7 from
114 http://wixtoolset.org and run dobuild.cmd from the
115 os/windows directory.
117 How to compile FIO on 64-bit Windows:
119 1. Install Cygwin (http://www.cygwin.com/setup.exe). Install 'make' and all
120 packages starting with 'mingw64-i686' and 'mingw64-x86_64'.
121 2. Download ftp://sourceware.org/pub/pthreads-win32/prebuilt-dll-2-9-1-release/dll/x64/pthreadGC2.dll
122 and copy to the fio source directory.
123 3. Open the Cygwin Terminal.
124 4. Go to the fio directory (source files).
133 --debug Enable some debugging options (see below)
134 --output Write output to file
135 --runtime Runtime in seconds
136 --latency-log Generate per-job latency logs
137 --bandwidth-log Generate per-job bandwidth logs
138 --minimal Minimal (terse) output
139 --output-format=type Output format (terse,json,normal)
140 --terse-version=type Terse version output format (default 3, or 2 or 4).
141 --version Print version info and exit
142 --help Print this page
143 --cpuclock-test Perform test/validation of CPU clock
144 --cmdhelp=cmd Print command help, "all" for all of them
145 --enghelp=engine Print ioengine help, or list available ioengines
146 --enghelp=engine,cmd Print help for an ioengine cmd
147 --showcmd Turn a job file into command line options
148 --readonly Turn on safety read-only checks, preventing
150 --eta=when When ETA estimate should be printed
151 May be "always", "never" or "auto"
152 --section=name Only run specified section in job file.
153 Multiple sections can be specified.
154 --alloc-size=kb Set smalloc pool to this size in kb (def 1024)
155 --warnings-fatal Fio parser warnings are fatal
156 --max-jobs Maximum number of threads/processes to support
157 --server=args Start backend server. See Client/Server section.
158 --client=host Connect to specified backend.
159 --idle-prof=option Report cpu idleness on a system or percpu basis
160 (option=system,percpu) or run unit work
161 calibration only (option=calibrate).
164 Any parameters following the options will be assumed to be job files,
165 unless they match a job file parameter. You can add as many as you want,
166 each job file will be regarded as a separate group and fio will stonewall
169 The --readonly switch is an extra safety guard to prevent accidentally
170 turning on a write setting when that is not desired. Fio will only write
171 if rw=write/randwrite/rw/randrw is given, but this extra safety net can
172 be used as an extra precaution. It will also enable a write check in the
173 io engine core to prevent an accidental write due to a fio bug.
175 The debug switch allows adding options that trigger certain logging
176 options in fio. Currently the options are:
178 process Dump info related to processes
179 file Dump info related to file actions
180 io Dump info related to IO queuing
181 mem Dump info related to memory allocations
182 blktrace Dump info related to blktrace setup
183 verify Dump info related to IO verification
184 all Enable all debug options
185 random Dump info related to random offset generation
186 parse Dump info related to option matching and parsing
187 diskutil Dump info related to disk utilization updates
188 job:x Dump info only related to job number x
189 mutex Dump info only related to mutex up/down ops
190 profile Dump info related to profile extensions
191 time Dump info related to internal time keeping
192 ? or help Show available debug options.
194 You can specify as many as you want, eg --debug=file,mem will enable
195 file and memory debugging.
197 The section switch is meant to make it easier to ship a bigger job file
198 instead of several smaller ones. Say you define a job file with light,
199 moderate, and heavy parts. Then you can ask fio to run the given part
200 only by giving it a --section=heavy command line option. The section
201 option only applies to job sections, the reserved 'global' section is
202 always parsed and taken into account.
204 Fio has an internal allocator for shared memory called smalloc. It
205 allocates shared structures from this pool. The pool defaults to 1024k
206 in size, and can grow to 128 pools. If running large jobs with randommap
207 enabled it can run out of memory, in which case the --alloc-size switch
208 is handy for starting with a larger pool size. The backing store is
209 files in /tmp. Fio cleans up after itself, while it is running you
210 may see .fio_smalloc.* files in /tmp.
216 See the HOWTO file for a more detailed description of parameters and what
217 they mean. This file contains the terse version. You can describe big and
218 complex setups with the command line, but generally it's a lot easier to
219 just write a simple job file to describe the workload. The job file format
220 is in the ini style format, as that is easy to read and write for the user.
222 The job file parameters are:
224 name=x Use 'x' as the identifier for this job.
225 description=x 'x' is a text description of the job.
226 directory=x Use 'x' as the top level directory for storing files
227 filename=x Force the use of 'x' as the filename for all files
228 in this thread. If not given, fio will make up
229 a suitable filename based on the thread and file
231 rw=x 'x' may be: read, randread, write, randwrite,
232 rw (read-write mix), randrw (read-write random mix)
233 rwmixcycle=x Base cycle for switching between read and write
235 rwmixread=x 'x' percentage of rw mix ios will be reads. If
236 rwmixwrite is also given, the last of the two will
237 be used if they don't add up to 100%.
238 rwmixwrite=x 'x' percentage of rw mix ios will be writes. See
240 rand_repeatable=x The sequence of random io blocks can be repeatable
241 across runs, if 'x' is 1.
242 size=x Set file size to x bytes (x string can include k/m/g)
243 ioengine=x 'x' may be: aio/libaio/linuxaio for Linux aio,
244 posixaio for POSIX aio, solarisaio for Solaris
245 native async IO, windowsaio for Windows native async IO,
246 sync for regular read/write io,
247 psync for regular pread/pwrite io, vsync for regular
248 readv/writev (with queuing emulation) mmap for mmap'ed
249 io, syslet-rw for syslet driven read/write, splice for
250 using splice/vmsplice, sg for direct SG_IO io, net
251 for network io, rdma for RDMA io, or cpuio for a
252 cycler burner load. sg only works on Linux on
253 SCSI (or SCSI-like devices, such as usb-storage or
254 sata/libata driven) devices. Fio also has a null
255 io engine, which is mainly used for testing
258 iodepth=x For async io, allow 'x' ios in flight
259 overwrite=x If 'x', layout a write file first.
260 nrfiles=x Spread io load over 'x' number of files per job,
262 prio=x Run io at prio X, 0-7 is the kernel allowed range
263 prioclass=x Run io at prio class X
264 bs=x Use 'x' for thread blocksize. May include k/m postfix.
265 bsrange=x-y Mix thread block sizes randomly between x and y. May
266 also include k/m postfix.
267 direct=x 1 for direct IO, 0 for buffered IO
268 thinktime=x "Think" x usec after each io
269 rate=x Throttle rate to x KB/sec
270 ratemin=x Quit if rate of x KB/sec can't be met
271 ratecycle=x ratemin averaged over x msecs
272 cpumask=x Only allow job to run on CPUs defined by mask.
273 cpus_allowed=x Like 'cpumask', but allow text setting of CPU affinity.
274 numa_cpu_nodes=x,y-z Allow job to run on specified NUMA nodes' CPU.
275 numa_mem_policy=m:x,y-z Setup numa memory allocation policy.
276 'm' stands for policy, such as local, interleave,
277 bind, prefer, local. 'x, y-z' are numa node(s) for
278 memory allocation according to policy.
279 fsync=x If writing with buffered IO, fsync after every
280 'x' blocks have been written.
281 end_fsync=x If 'x', run fsync() after end-of-job.
282 startdelay=x Start this thread x seconds after startup
283 runtime=x Terminate x seconds after startup. Can include a
284 normal time suffix if not given in seconds, such as
285 'm' for minutes, 'h' for hours, and 'd' for days.
286 offset=x Start io at offset x (x string can include k/m/g)
287 invalidate=x Invalidate page cache for file prior to doing io
288 sync=x Use sync writes if x and writing buffered IO.
289 mem=x If x == malloc, use malloc for buffers. If x == shm,
290 use shared memory for buffers. If x == mmap, use
292 exitall When one thread quits, terminate the others
293 bwavgtime=x Average bandwidth stats over an x msec window.
294 create_serialize=x If 'x', serialize file creation.
295 create_fsync=x If 'x', run fsync() after file creation.
296 unlink If set, unlink files when done.
297 loops=x Run the job 'x' number of times.
298 verify=x If 'x' == md5, use md5 for verifies. If 'x' == crc32,
299 use crc32 for verifies. md5 is 'safer', but crc32 is
300 a lot faster. Only makes sense for writing to a file.
301 For other types of checksumming, see HOWTO.
302 stonewall Wait for preceeding jobs to end before running.
303 numjobs=x Create 'x' similar entries for this job
304 thread Use pthreads instead of forked jobs
306 zoneskip=y Zone options must be paired. If given, the job
307 will skip y bytes for every x read/written. This
308 can be used to gauge hard drive speed over the entire
309 platter, without reading everything. Both x/y can
310 include k/m/g suffix.
311 read_iolog=x Open and read io pattern from file 'x'. The file format
312 is described in the HOWTO.
313 write_iolog=x Write an iolog to file 'x' in the same format as iolog.
314 The iolog options are exclusive, if both given the
315 read iolog will be performed. Specify a separate file
316 for each job, otherwise the iologs will be interspersed
317 and the file may be corrupt.
318 write_bw_log Write a bandwidth log.
319 write_lat_log Write a latency log.
320 lockmem=x Lock down x amount of memory on the machine, to
321 simulate a machine with less memory available. x can
322 include k/m/g suffix.
323 nice=x Run job at given nice value.
324 exec_prerun=x Run 'x' before job io is begun.
325 exec_postrun=x Run 'x' after job io has finished.
326 ioscheduler=x Use ioscheduler 'x' for this job.
327 cpuload=x For a CPU io thread, percentage of CPU time to attempt
329 cpuchunks=x Split burn cycles into pieces of x usecs.
336 Normally you would run fio as a stand-alone application on the machine
337 where the IO workload should be generated. However, it is also possible to
338 run the frontend and backend of fio separately. This makes it possible to
339 have a fio server running on the machine(s) where the IO workload should
340 be running, while controlling it from another machine.
342 To start the server, you would do:
346 on that machine, where args defines what fio listens to. The arguments
347 are of the form 'type,hostname or IP,port'. 'type' is either 'ip' (or ip4)
348 for TCP/IP v4, 'ip6' for TCP/IP v6, or 'sock' for a local unix domain socket.
349 'hostname' is either a hostname or IP address, and 'port' is the port to
350 listen to (only valid for TCP/IP, not a local socket). Some examples:
354 Start a fio server, listening on all interfaces on the default port (8765).
356 2) fio --server=ip:hostname,4444
358 Start a fio server, listening on IP belonging to hostname and on port 4444.
360 3) fio --server=ip6:::1,4444
362 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
364 4) fio --server=,4444
366 Start a fio server, listening on all interfaces on port 4444.
368 5) fio --server=1.2.3.4
370 Start a fio server, listening on IP 1.2.3.4 on the default port.
372 6) fio --server=sock:/tmp/fio.sock
374 Start a fio server, listening on the local socket /tmp/fio.sock.
376 When a server is running, you can connect to it from a client. The client
379 fio --local-args --client=server --remote-args <job file(s)>
381 where --local-args are arguments that are local to the client where it is
382 running, 'server' is the connect string, and --remote-args and <job file(s)>
383 are sent to the server. The 'server' string follows the same format as it
384 does on the server side, to allow IP/hostname/socket and port strings.
385 You can connect to multiple clients as well, to do that you could run:
387 fio --client=server2 <job file(s)> --client=server2 <job file(s)>
393 Fio works on (at least) Linux, Solaris, AIX, HP-UX, OSX, NetBSD, Windows
394 and FreeBSD. Some features and/or options may only be available on some of
395 the platforms, typically because those features only apply to that platform
396 (like the solarisaio engine, or the splice engine on Linux).
398 Some features are not available on FreeBSD/Solaris even if they could be
399 implemented, I'd be happy to take patches for that. An example of that is
400 disk utility statistics and (I think) huge page support, support for that
401 does exist in FreeBSD/Solaris.
403 Fio uses pthread mutexes for signalling and locking and FreeBSD does not
404 support process shared pthread mutexes. As a result, only threads are
405 supported on FreeBSD. This could be fixed with sysv ipc locking or
406 other locking alternatives.
408 Other *BSD platforms are untested, but fio should work there almost out
409 of the box. Since I don't do test runs or even compiles on those platforms,
410 your mileage may vary. Sending me patches for other platforms is greatly
411 appreciated. There's a lot of value in having the same test/benchmark tool
412 available on all platforms.
414 Note that POSIX aio is not enabled by default on AIX. If you get messages like:
416 Symbol resolution failed for /usr/lib/libc.a(posix_aio.o) because:
417 Symbol _posix_kaio_rdwr (number 2) is not exported from dependent module /unix.
419 you need to enable POSIX aio. Run the following commands as root:
421 # lsdev -C -l posix_aio0
422 posix_aio0 Defined Posix Asynchronous I/O
423 # cfgmgr -l posix_aio0
424 # lsdev -C -l posix_aio0
425 posix_aio0 Available Posix Asynchronous I/O
427 POSIX aio should work now. To make the change permanent:
429 # chdev -l posix_aio0 -P -a autoconfig='available'
436 Fio was written by Jens Axboe <axboe@kernel.dk> to enable flexible testing
437 of the Linux IO subsystem and schedulers. He got tired of writing
438 specific test applications to simulate a given workload, and found that
439 the existing io benchmark/test tools out there weren't flexible enough
440 to do what he wanted.
442 Jens Axboe <axboe@kernel.dk> 20060905