README: add hint on missing libaio-devel package
[fio.git] / README
CommitLineData
ebac4655
JA
1fio
2---
3
79809113
JA
4fio is a tool that will spawn a number of threads or processes doing a
5particular type of io action as specified by the user. fio takes a
6number of global parameters, each inherited by the thread unless
7otherwise parameters given to them overriding that setting is given.
8The typical use of fio is to write a job file matching the io load
9one wants to simulate.
ebac4655 10
2b02b546
JA
11
12Source
13------
14
15fio resides in a git repo, the canonical place is:
16
6b3eccb1 17git://git.kernel.dk/fio.git
97f049c9
JA
18
19The http protocol also works, path is the same.
2b02b546 20
79809113
JA
21Snapshots are frequently generated and they include the git meta data as
22well. You can download them here:
2b02b546
JA
23
24http://brick.kernel.dk/snaps/
25
a68594cb
JA
26Pascal Bleser <guru@unixtech.be> has fio RPMs in his repository for
27SUSE variants, you can find them here:
1053a106
JA
28
29http://linux01.gwdg.de/~pbleser/rpm-navigation.php?cat=System/fio
30
a68594cb
JA
31Dag Wieƫrs has RPMs for Red Hat related distros, find them here:
32
33http://dag.wieers.com/rpm/packages/fio/
34
244e170e
JA
35Mandriva has integrated fio into their package repository, so installing
36on that distro should be as easy as typing 'urpmi fio'.
37
2b02b546 38
726f6ff0
JA
39Mailing list
40------------
41
42There's a mailing list associated with fio. It's meant for general
43discussion, bug reporting, questions - basically anything that has to
44do with fio. An automated mail detailing recent commits is automatically
45sent to the list at most daily. The list address is fio-devel@kernel.dk,
46subscribe by sending an empty email to fio-devel+subscribe@kernel.dk.
47
48
bbfd6b00
JA
49Building
50--------
51
52Just type 'make' and 'make install'. If on FreeBSD, for now you have to
53specify the FreeBSD Makefile with -f, eg:
54
55$ make -f Makefile.Freebsd && make -f Makefile.FreeBSD install
56
edffcb96 57Likewise with OpenSolaris, use the Makefile.solaris to compile there.
bbfd6b00
JA
58This might change in the future if I opt for an autoconf type setup.
59
6de43c1b
JA
60If your compile fails with an error like this:
61
62 CC gettime.o
63In file included from fio.h:23,
64 from gettime.c:8:
65os/os.h:15:20: error: libaio.h: No such file or directory
66In file included from gettime.c:8:
67fio.h:119: error: field 'iocb' has incomplete type
68make: *** [gettime.o] Error 1
69
70Check that you have the libaio development package installed. On RPM
71based distros, it's typically called libaio-devel.
72
bbfd6b00 73
972cfd25
JA
74Command line
75------------
ebac4655
JA
76
77$ fio
ee56ad50 78 --debug Enable some debugging options (see below)
b4692828 79 --output Write output to file
062c6022 80 --timeout Runtime in seconds
b4692828
JA
81 --latency-log Generate per-job latency logs
82 --bandwidth-log Generate per-job bandwidth logs
83 --minimal Minimal (terse) output
84 --version Print version info and exit
fd28ca49
JA
85 --help Print this page
86 --cmdhelp=cmd Print command help, "all" for all of them
cca73aa7 87 --showcmd Turn a job file into command line options
062c6022 88 --readonly Turn on safety read-only checks, preventing writes
e592a06b
AC
89 --eta=when When ETA estimate should be printed
90 May be "always", "never" or "auto"
01f06b63 91 --section=name Only run specified section in job file
2b386d25 92 --alloc-size=kb Set smalloc pool to this size in kb (def 1024)
e592a06b 93
b4692828
JA
94
95Any parameters following the options will be assumed to be job files,
96unless they match a job file parameter. You can add as many as you want,
97each job file will be regarded as a separate group and fio will stonewall
98its execution.
972cfd25 99
724e4435
JA
100The --readonly switch is an extra safety guard to prevent accidentically
101turning on a write setting when that is not desired. Fio will only write
102if rw=write/randwrite/rw/randrw is given, but this extra safety net can
103be used as an extra precaution. It will also enable a write check in the
104io engine core to prevent an accidental write due to a fio bug.
105
ee56ad50
JA
106The debug switch allows adding options that trigger certain logging
107options in fio. Currently the options are:
108
109 process Dump info related to processes
110 file Dump info related to file actions
111 io Dump info related to IO queuing
112 mem Dump info related to memory allocations
bd6f78b2
JA
113 blktrace Dump info related to blktrace setup
114 verify Dump info related to IO verification
115 all Enable all debug options
811a0d06 116 random Dump info related to random offset generation
a3d741fa 117 parse Dump info related to option matching and parsing
cd991b9e 118 diskutil Dump info related to disk utilization updates
bd6f78b2 119 ? or help Show available debug options.
ee56ad50
JA
120
121You can specify as many as you want, eg --debug=file,mem will enable
bd6f78b2 122file and memory debugging.
ee56ad50 123
01f06b63
JA
124The section switch is meant to make it easier to ship a bigger job file
125instead of several smaller ones. Say you define a job file with light,
126moderate, and heavy parts. Then you can ask fio to run the given part
127only by giving it a --section=heavy command line option. The section
128option only applies to job sections, the reserved 'global' section is
129always parsed and taken into account.
130
2b386d25
JA
131Fio has an internal allocator for shared memory called smalloc. It
132allocates shared structures from this pool. The pool defaults to 1024k
133in size, and can grow to 32 pools. If running large jobs with randommap
134enabled it can run out of memory, in which case the --alloc-size switch
135is handy for starting with a larger pool size.
136
79809113
JA
137
138Job file
139--------
140
71bfa161 141See the HOWTO file for a more detailed description of parameters and what
4661f3d0
JA
142they mean. This file contains the terse version. You can describe big and
143complex setups with the command line, but generally it's a lot easier to
71bfa161 144just write a simple job file to describe the workload. The job file format
4661f3d0 145is in the ini style format, as that is easy to read and write for the user.
79809113
JA
146
147The job file parameters are:
ebac4655 148
01452055 149 name=x Use 'x' as the identifier for this job.
61697c37 150 description=x 'x' is a text description of the job.
ebac4655 151 directory=x Use 'x' as the top level directory for storing files
b50b8755
JA
152 filename=x Force the use of 'x' as the filename for all files
153 in this thread. If not given, fio will make up
154 a suitable filename based on the thread and file
155 number.
3d60d1ed
JA
156 rw=x 'x' may be: read, randread, write, randwrite,
157 rw (read-write mix), randrw (read-write random mix)
a6ccc7be
JA
158 rwmixcycle=x Base cycle for switching between read and write
159 in msecs.
160 rwmixread=x 'x' percentage of rw mix ios will be reads. If
161 rwmixwrite is also given, the last of the two will
162 be used if they don't add up to 100%.
163 rwmixwrite=x 'x' percentage of rw mix ios will be writes. See
164 rwmixread.
9ebc27e1
JA
165 rand_repeatable=x The sequence of random io blocks can be repeatable
166 across runs, if 'x' is 1.
ebac4655
JA
167 size=x Set file size to x bytes (x string can include k/m/g)
168 ioengine=x 'x' may be: aio/libaio/linuxaio for Linux aio,
169 posixaio for POSIX aio, sync for regular read/write io,
1d2af02a
JA
170 psync for regular pread/pwrite io, vsync for regular
171 readv/writev (with queuing emulation) mmap for mmap'ed
172 io, syslet-rw for syslet driven read/write, splice for
173 using splice/vmsplice, sgio for direct SG_IO io, net
174 for network io, or cpuio for a cycler burner load. sgio
175 only works on Linux on SCSI (or SCSI-like devices, such
176 as usb-storage or sata/libata driven) devices. Fio also
177 has a null io engine, which is mainly used for testing
178 fio itself.
179
ebac4655
JA
180 iodepth=x For async io, allow 'x' ios in flight
181 overwrite=x If 'x', layout a write file first.
53cdc686
JA
182 nrfiles=x Spread io load over 'x' number of files per job,
183 if possible.
ebac4655
JA
184 prio=x Run io at prio X, 0-7 is the kernel allowed range
185 prioclass=x Run io at prio class X
186 bs=x Use 'x' for thread blocksize. May include k/m postfix.
187 bsrange=x-y Mix thread block sizes randomly between x and y. May
188 also include k/m postfix.
189 direct=x 1 for direct IO, 0 for buffered IO
190 thinktime=x "Think" x usec after each io
191 rate=x Throttle rate to x KiB/sec
192 ratemin=x Quit if rate of x KiB/sec can't be met
193 ratecycle=x ratemin averaged over x msecs
194 cpumask=x Only allow job to run on CPUs defined by mask.
d2e268b0 195 cpus_allowed=x Like 'cpumask', but allow text setting of CPU affinity.
795407ca
JA
196 fsync=x If writing with buffered IO, fsync after every
197 'x' blocks have been written.
198 end_fsync=x If 'x', run fsync() after end-of-job.
ebac4655 199 startdelay=x Start this thread x seconds after startup
03b74b3e 200 runtime=x Terminate x seconds after startup. Can include a
906c8d75
JA
201 normal time suffix if not given in seconds, such as
202 'm' for minutes, 'h' for hours, and 'd' for days.
ebac4655
JA
203 offset=x Start io at offset x (x string can include k/m/g)
204 invalidate=x Invalidate page cache for file prior to doing io
795407ca 205 sync=x Use sync writes if x and writing buffered IO.
ebac4655 206 mem=x If x == malloc, use malloc for buffers. If x == shm,
795407ca
JA
207 use shared memory for buffers. If x == mmap, use
208 anonymous mmap.
ebac4655
JA
209 exitall When one thread quits, terminate the others
210 bwavgtime=x Average bandwidth stats over an x msec window.
211 create_serialize=x If 'x', serialize file creation.
212 create_fsync=x If 'x', run fsync() after file creation.
f6cbb269 213 unlink If set, unlink files when done.
ebac4655
JA
214 loops=x Run the job 'x' number of times.
215 verify=x If 'x' == md5, use md5 for verifies. If 'x' == crc32,
216 use crc32 for verifies. md5 is 'safer', but crc32 is
217 a lot faster. Only makes sense for writing to a file.
218 stonewall Wait for preceeding jobs to end before running.
219 numjobs=x Create 'x' similar entries for this job
220 thread Use pthreads instead of forked jobs
20dc95c4
JA
221 zonesize=x
222 zoneskip=y Zone options must be paired. If given, the job
223 will skip y bytes for every x read/written. This
224 can be used to gauge hard drive speed over the entire
225 platter, without reading everything. Both x/y can
226 include k/m/g suffix.
aea47d44
JA
227 iolog=x Open and read io pattern from file 'x'. The file must
228 contain one io action per line in the following format:
229 rw, offset, length
230 where with rw=0/1 for read/write, and the offset
231 and length entries being in bytes.
843a7413
JA
232 write_iolog=x Write an iolog to file 'x' in the same format as iolog.
233 The iolog options are exclusive, if both given the
234 read iolog will be performed.
ec94ec56
JA
235 write_bw_log Write a bandwidth log.
236 write_lat_log Write a latency log.
c04f7ec3
JA
237 lockmem=x Lock down x amount of memory on the machine, to
238 simulate a machine with less memory available. x can
239 include k/m/g suffix.
b6f4d880 240 nice=x Run job at given nice value.
4e0ba8af
JA
241 exec_prerun=x Run 'x' before job io is begun.
242 exec_postrun=x Run 'x' after job io has finished.
da86774e 243 ioscheduler=x Use ioscheduler 'x' for this job.
b990b5c0
JA
244 cpuload=x For a CPU io thread, percentage of CPU time to attempt
245 to burn.
ba0fbe10 246 cpuchunks=x Split burn cycles into pieces of x usecs.
ebac4655 247
79809113 248
79809113
JA
249Author
250------
251
aae22ca7 252Fio was written by Jens Axboe <axboe@kernel.dk> to enable flexible testing
79809113
JA
253of the Linux IO subsystem and schedulers. He got tired of writing
254specific test applications to simulate a given workload, and found that
255the existing io benchmark/test tools out there weren't flexible enough
256to do what he wanted.
257
aae22ca7 258Jens Axboe <axboe@kernel.dk> 20060905
79809113 259