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 and use gmake (not make), eg:
55 $ gmake -f Makefile.Freebsd && gmake -f Makefile.FreeBSD install
57 Likewise with OpenSolaris, use the Makefile.solaris to compile there.
58 The OpenSolaris make should work fine. This might change in the
59 future if I opt for an autoconf type setup.
61 If your compile fails with an error like this:
64 In file included from fio.h:23,
66 os/os.h:15:20: error: libaio.h: No such file or directory
67 In file included from gettime.c:8:
68 fio.h:119: error: field 'iocb' has incomplete type
69 make: *** [gettime.o] Error 1
71 Check that you have the libaio development package installed. On RPM
72 based distros, it's typically called libaio-devel.
79 --debug Enable some debugging options (see below)
80 --output Write output to file
81 --timeout Runtime in seconds
82 --latency-log Generate per-job latency logs
83 --bandwidth-log Generate per-job bandwidth logs
84 --minimal Minimal (terse) output
85 --version Print version info and exit
86 --help Print this page
87 --cmdhelp=cmd Print command help, "all" for all of them
88 --showcmd Turn a job file into command line options
89 --readonly Turn on safety read-only checks, preventing writes
90 --eta=when When ETA estimate should be printed
91 May be "always", "never" or "auto"
92 --section=name Only run specified section in job file
93 --alloc-size=kb Set smalloc pool to this size in kb (def 1024)
96 Any parameters following the options will be assumed to be job files,
97 unless they match a job file parameter. You can add as many as you want,
98 each job file will be regarded as a separate group and fio will stonewall
101 The --readonly switch is an extra safety guard to prevent accidentically
102 turning on a write setting when that is not desired. Fio will only write
103 if rw=write/randwrite/rw/randrw is given, but this extra safety net can
104 be used as an extra precaution. It will also enable a write check in the
105 io engine core to prevent an accidental write due to a fio bug.
107 The debug switch allows adding options that trigger certain logging
108 options in fio. Currently the options are:
110 process Dump info related to processes
111 file Dump info related to file actions
112 io Dump info related to IO queuing
113 mem Dump info related to memory allocations
114 blktrace Dump info related to blktrace setup
115 verify Dump info related to IO verification
116 all Enable all debug options
117 random Dump info related to random offset generation
118 parse Dump info related to option matching and parsing
119 diskutil Dump info related to disk utilization updates
120 job:x Dump info only related to job number x
121 ? or help Show available debug options.
123 You can specify as many as you want, eg --debug=file,mem will enable
124 file and memory debugging.
126 The section switch is meant to make it easier to ship a bigger job file
127 instead of several smaller ones. Say you define a job file with light,
128 moderate, and heavy parts. Then you can ask fio to run the given part
129 only by giving it a --section=heavy command line option. The section
130 option only applies to job sections, the reserved 'global' section is
131 always parsed and taken into account.
133 Fio has an internal allocator for shared memory called smalloc. It
134 allocates shared structures from this pool. The pool defaults to 1024k
135 in size, and can grow to 32 pools. If running large jobs with randommap
136 enabled it can run out of memory, in which case the --alloc-size switch
137 is handy for starting with a larger pool size.
143 See the HOWTO file for a more detailed description of parameters and what
144 they mean. This file contains the terse version. You can describe big and
145 complex setups with the command line, but generally it's a lot easier to
146 just write a simple job file to describe the workload. The job file format
147 is in the ini style format, as that is easy to read and write for the user.
149 The job file parameters are:
151 name=x Use 'x' as the identifier for this job.
152 description=x 'x' is a text description of the job.
153 directory=x Use 'x' as the top level directory for storing files
154 filename=x Force the use of 'x' as the filename for all files
155 in this thread. If not given, fio will make up
156 a suitable filename based on the thread and file
158 rw=x 'x' may be: read, randread, write, randwrite,
159 rw (read-write mix), randrw (read-write random mix)
160 rwmixcycle=x Base cycle for switching between read and write
162 rwmixread=x 'x' percentage of rw mix ios will be reads. If
163 rwmixwrite is also given, the last of the two will
164 be used if they don't add up to 100%.
165 rwmixwrite=x 'x' percentage of rw mix ios will be writes. See
167 rand_repeatable=x The sequence of random io blocks can be repeatable
168 across runs, if 'x' is 1.
169 size=x Set file size to x bytes (x string can include k/m/g)
170 ioengine=x 'x' may be: aio/libaio/linuxaio for Linux aio,
171 posixaio for POSIX aio, solarisaio for Solaris
172 native async IO, sync for regular read/write io,
173 psync for regular pread/pwrite io, vsync for regular
174 readv/writev (with queuing emulation) mmap for mmap'ed
175 io, syslet-rw for syslet driven read/write, splice for
176 using splice/vmsplice, sg for direct SG_IO io, net
177 for network io, or cpuio for a cycler burner load. sg
178 only works on Linux on SCSI (or SCSI-like devices, such
179 as usb-storage or sata/libata driven) devices. Fio also
180 has a null io engine, which is mainly used for testing
183 iodepth=x For async io, allow 'x' ios in flight
184 overwrite=x If 'x', layout a write file first.
185 nrfiles=x Spread io load over 'x' number of files per job,
187 prio=x Run io at prio X, 0-7 is the kernel allowed range
188 prioclass=x Run io at prio class X
189 bs=x Use 'x' for thread blocksize. May include k/m postfix.
190 bsrange=x-y Mix thread block sizes randomly between x and y. May
191 also include k/m postfix.
192 direct=x 1 for direct IO, 0 for buffered IO
193 thinktime=x "Think" x usec after each io
194 rate=x Throttle rate to x KiB/sec
195 ratemin=x Quit if rate of x KiB/sec can't be met
196 ratecycle=x ratemin averaged over x msecs
197 cpumask=x Only allow job to run on CPUs defined by mask.
198 cpus_allowed=x Like 'cpumask', but allow text setting of CPU affinity.
199 fsync=x If writing with buffered IO, fsync after every
200 'x' blocks have been written.
201 end_fsync=x If 'x', run fsync() after end-of-job.
202 startdelay=x Start this thread x seconds after startup
203 runtime=x Terminate x seconds after startup. Can include a
204 normal time suffix if not given in seconds, such as
205 'm' for minutes, 'h' for hours, and 'd' for days.
206 offset=x Start io at offset x (x string can include k/m/g)
207 invalidate=x Invalidate page cache for file prior to doing io
208 sync=x Use sync writes if x and writing buffered IO.
209 mem=x If x == malloc, use malloc for buffers. If x == shm,
210 use shared memory for buffers. If x == mmap, use
212 exitall When one thread quits, terminate the others
213 bwavgtime=x Average bandwidth stats over an x msec window.
214 create_serialize=x If 'x', serialize file creation.
215 create_fsync=x If 'x', run fsync() after file creation.
216 unlink If set, unlink files when done.
217 loops=x Run the job 'x' number of times.
218 verify=x If 'x' == md5, use md5 for verifies. If 'x' == crc32,
219 use crc32 for verifies. md5 is 'safer', but crc32 is
220 a lot faster. Only makes sense for writing to a file.
221 stonewall Wait for preceeding jobs to end before running.
222 numjobs=x Create 'x' similar entries for this job
223 thread Use pthreads instead of forked jobs
225 zoneskip=y Zone options must be paired. If given, the job
226 will skip y bytes for every x read/written. This
227 can be used to gauge hard drive speed over the entire
228 platter, without reading everything. Both x/y can
229 include k/m/g suffix.
230 iolog=x Open and read io pattern from file 'x'. The file must
231 contain one io action per line in the following format:
233 where with rw=0/1 for read/write, and the offset
234 and length entries being in bytes.
235 write_iolog=x Write an iolog to file 'x' in the same format as iolog.
236 The iolog options are exclusive, if both given the
237 read iolog will be performed.
238 write_bw_log Write a bandwidth log.
239 write_lat_log Write a latency log.
240 lockmem=x Lock down x amount of memory on the machine, to
241 simulate a machine with less memory available. x can
242 include k/m/g suffix.
243 nice=x Run job at given nice value.
244 exec_prerun=x Run 'x' before job io is begun.
245 exec_postrun=x Run 'x' after job io has finished.
246 ioscheduler=x Use ioscheduler 'x' for this job.
247 cpuload=x For a CPU io thread, percentage of CPU time to attempt
249 cpuchunks=x Split burn cycles into pieces of x usecs.
255 Fio was written by Jens Axboe <axboe@kernel.dk> to enable flexible testing
256 of the Linux IO subsystem and schedulers. He got tired of writing
257 specific test applications to simulate a given workload, and found that
258 the existing io benchmark/test tools out there weren't flexible enough
259 to do what he wanted.
261 Jens Axboe <axboe@kernel.dk> 20060905