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/data/git/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/
39 There's a mailing list associated with fio. It's meant for general
40 discussion, bug reporting, questions - basically anything that has to
41 do with fio. An automated mail detailing recent commits is automatically
42 sent to the list at most daily. The list address is fio-devel@kernel.dk,
43 subscribe by sending an empty email to fio-devel+subscribe@kernel.dk.
49 Just type 'make' and 'make install'. If on FreeBSD, for now you have to
50 specify the FreeBSD Makefile with -f, eg:
52 $ make -f Makefile.Freebsd && make -f Makefile.FreeBSD install
54 Likewise with OpenSolaris, use the Makefile.solaris to compile there.
55 This might change in the future if I opt for an autoconf type setup.
62 --output Write output to file
63 --runtime Runtime in seconds
64 --latency-log Generate per-job latency logs
65 --bandwidth-log Generate per-job bandwidth logs
66 --minimal Minimal (terse) output
67 --version Print version info and exit
68 --help Print this page
69 --cmdhelp=cmd Print command help, "all" for all of them
70 --showcmd Turn a job file into command line options
72 Any parameters following the options will be assumed to be job files,
73 unless they match a job file parameter. You can add as many as you want,
74 each job file will be regarded as a separate group and fio will stonewall
81 See the HOWTO file for a more detailed description of parameters and what
82 they mean. This file contains the terse version. Only a few options can
83 be controlled with command line parameters, generally it's a lot easier to
84 just write a simple job file to describe the workload. The job file format
85 is in the ini style format, as it's easy to read and write for the user.
87 The job file parameters are:
89 name=x Use 'x' as the identifier for this job.
90 description=x 'x' is a text description of the job.
91 directory=x Use 'x' as the top level directory for storing files
92 filename=x Force the use of 'x' as the filename for all files
93 in this thread. If not given, fio will make up
94 a suitable filename based on the thread and file
96 rw=x 'x' may be: read, randread, write, randwrite,
97 rw (read-write mix), randrw (read-write random mix)
98 rwmixcycle=x Base cycle for switching between read and write
100 rwmixread=x 'x' percentage of rw mix ios will be reads. If
101 rwmixwrite is also given, the last of the two will
102 be used if they don't add up to 100%.
103 rwmixwrite=x 'x' percentage of rw mix ios will be writes. See
105 rand_repeatable=x The sequence of random io blocks can be repeatable
106 across runs, if 'x' is 1.
107 size=x Set file size to x bytes (x string can include k/m/g)
108 ioengine=x 'x' may be: aio/libaio/linuxaio for Linux aio,
109 posixaio for POSIX aio, sync for regular read/write io,
110 mmap for mmap'ed io, syslet-rw for syslet driven
111 read/write, splice for using splice/vmsplice,
112 sgio for direct SG_IO io, net for network io, or cpuio
113 for a cycler burner load. sgio only works on Linux on
114 SCSI (or SCSI-like devices, such as usb-storage or
115 sata/libata driven) devices. Fio also has a null io
116 engine, which is mainly used for testing fio itself.
117 iodepth=x For async io, allow 'x' ios in flight
118 overwrite=x If 'x', layout a write file first.
119 nrfiles=x Spread io load over 'x' number of files per job,
121 prio=x Run io at prio X, 0-7 is the kernel allowed range
122 prioclass=x Run io at prio class X
123 bs=x Use 'x' for thread blocksize. May include k/m postfix.
124 bsrange=x-y Mix thread block sizes randomly between x and y. May
125 also include k/m postfix.
126 direct=x 1 for direct IO, 0 for buffered IO
127 thinktime=x "Think" x usec after each io
128 rate=x Throttle rate to x KiB/sec
129 ratemin=x Quit if rate of x KiB/sec can't be met
130 ratecycle=x ratemin averaged over x msecs
131 cpumask=x Only allow job to run on CPUs defined by mask.
132 fsync=x If writing with buffered IO, fsync after every
133 'x' blocks have been written.
134 end_fsync=x If 'x', run fsync() after end-of-job.
135 startdelay=x Start this thread x seconds after startup
136 runtime=x Terminate x seconds after startup. Can include a
137 normal time suffix if not given in seconds, such as
138 'm' for minutes, 'h' for hours, and 'd' for days.
139 offset=x Start io at offset x (x string can include k/m/g)
140 invalidate=x Invalidate page cache for file prior to doing io
141 sync=x Use sync writes if x and writing buffered IO.
142 mem=x If x == malloc, use malloc for buffers. If x == shm,
143 use shared memory for buffers. If x == mmap, use
145 exitall When one thread quits, terminate the others
146 bwavgtime=x Average bandwidth stats over an x msec window.
147 create_serialize=x If 'x', serialize file creation.
148 create_fsync=x If 'x', run fsync() after file creation.
149 unlink If set, unlink files when done.
150 loops=x Run the job 'x' number of times.
151 verify=x If 'x' == md5, use md5 for verifies. If 'x' == crc32,
152 use crc32 for verifies. md5 is 'safer', but crc32 is
153 a lot faster. Only makes sense for writing to a file.
154 stonewall Wait for preceeding jobs to end before running.
155 numjobs=x Create 'x' similar entries for this job
156 thread Use pthreads instead of forked jobs
158 zoneskip=y Zone options must be paired. If given, the job
159 will skip y bytes for every x read/written. This
160 can be used to gauge hard drive speed over the entire
161 platter, without reading everything. Both x/y can
162 include k/m/g suffix.
163 iolog=x Open and read io pattern from file 'x'. The file must
164 contain one io action per line in the following format:
166 where with rw=0/1 for read/write, and the offset
167 and length entries being in bytes.
168 write_iolog=x Write an iolog to file 'x' in the same format as iolog.
169 The iolog options are exclusive, if both given the
170 read iolog will be performed.
171 write_bw_log Write a bandwidth log.
172 write_lat_log Write a latency log.
173 lockmem=x Lock down x amount of memory on the machine, to
174 simulate a machine with less memory available. x can
175 include k/m/g suffix.
176 nice=x Run job at given nice value.
177 exec_prerun=x Run 'x' before job io is begun.
178 exec_postrun=x Run 'x' after job io has finished.
179 ioscheduler=x Use ioscheduler 'x' for this job.
180 cpuload=x For a CPU io thread, percentage of CPU time to attempt
182 cpuchunks=x Split burn cycles into pieces of x usecs.
188 Fio was written by Jens Axboe <axboe@kernel.dk> to enable flexible testing
189 of the Linux IO subsystem and schedulers. He got tired of writing
190 specific test applications to simulate a given workload, and found that
191 the existing io benchmark/test tools out there weren't flexible enough
192 to do what he wanted.
194 Jens Axboe <axboe@kernel.dk> 20060905