Add intel hardware assisted crc32c support
[fio.git] / HOWTO
CommitLineData
71bfa161
JA
1Table of contents
2-----------------
3
41. Overview
52. How fio works
63. Running fio
74. Job file format
85. Detailed list of parameters
96. Normal output
107. Terse output
11
12
131.0 Overview and history
14------------------------
15fio was originally written to save me the hassle of writing special test
16case programs when I wanted to test a specific workload, either for
17performance reasons or to find/reproduce a bug. The process of writing
18such a test app can be tiresome, especially if you have to do it often.
19Hence I needed a tool that would be able to simulate a given io workload
20without resorting to writing a tailored test case again and again.
21
22A test work load is difficult to define, though. There can be any number
23of processes or threads involved, and they can each be using their own
24way of generating io. You could have someone dirtying large amounts of
25memory in an memory mapped file, or maybe several threads issuing
26reads using asynchronous io. fio needed to be flexible enough to
27simulate both of these cases, and many more.
28
292.0 How fio works
30-----------------
31The first step in getting fio to simulate a desired io workload, is
32writing a job file describing that specific setup. A job file may contain
33any number of threads and/or files - the typical contents of the job file
34is a global section defining shared parameters, and one or more job
35sections describing the jobs involved. When run, fio parses this file
36and sets everything up as described. If we break down a job from top to
37bottom, it contains the following basic parameters:
38
39 IO type Defines the io pattern issued to the file(s).
40 We may only be reading sequentially from this
41 file(s), or we may be writing randomly. Or even
42 mixing reads and writes, sequentially or randomly.
43
44 Block size In how large chunks are we issuing io? This may be
45 a single value, or it may describe a range of
46 block sizes.
47
48 IO size How much data are we going to be reading/writing.
49
50 IO engine How do we issue io? We could be memory mapping the
51 file, we could be using regular read/write, we
d0ff85df 52 could be using splice, async io, syslet, or even
71bfa161
JA
53 SG (SCSI generic sg).
54
6c219763 55 IO depth If the io engine is async, how large a queuing
71bfa161
JA
56 depth do we want to maintain?
57
58 IO type Should we be doing buffered io, or direct/raw io?
59
60 Num files How many files are we spreading the workload over.
61
62 Num threads How many threads or processes should we spread
63 this workload over.
64
65The above are the basic parameters defined for a workload, in addition
66there's a multitude of parameters that modify other aspects of how this
67job behaves.
68
69
703.0 Running fio
71---------------
72See the README file for command line parameters, there are only a few
73of them.
74
75Running fio is normally the easiest part - you just give it the job file
76(or job files) as parameters:
77
78$ fio job_file
79
80and it will start doing what the job_file tells it to do. You can give
81more than one job file on the command line, fio will serialize the running
82of those files. Internally that is the same as using the 'stonewall'
83parameter described the the parameter section.
84
b4692828
JA
85If the job file contains only one job, you may as well just give the
86parameters on the command line. The command line parameters are identical
87to the job parameters, with a few extra that control global parameters
88(see README). For example, for the job file parameter iodepth=2, the
c2b1e753
JA
89mirror command line option would be --iodepth 2 or --iodepth=2. You can
90also use the command line for giving more than one job entry. For each
91--name option that fio sees, it will start a new job with that name.
92Command line entries following a --name entry will apply to that job,
93until there are no more entries or a new --name entry is seen. This is
94similar to the job file options, where each option applies to the current
95job until a new [] job entry is seen.
b4692828 96
71bfa161
JA
97fio does not need to run as root, except if the files or devices specified
98in the job section requires that. Some other options may also be restricted,
6c219763 99such as memory locking, io scheduler switching, and decreasing the nice value.
71bfa161
JA
100
101
1024.0 Job file format
103-------------------
104As previously described, fio accepts one or more job files describing
105what it is supposed to do. The job file format is the classic ini file,
106where the names enclosed in [] brackets define the job name. You are free
107to use any ascii name you want, except 'global' which has special meaning.
108A global section sets defaults for the jobs described in that file. A job
109may override a global section parameter, and a job file may even have
110several global sections if so desired. A job is only affected by a global
65db0851
JA
111section residing above it. If the first character in a line is a ';' or a
112'#', the entire line is discarded as a comment.
71bfa161
JA
113
114So lets look at a really simple job file that define to threads, each
115randomly reading from a 128MiB file.
116
117; -- start job file --
118[global]
119rw=randread
120size=128m
121
122[job1]
123
124[job2]
125
126; -- end job file --
127
128As you can see, the job file sections themselves are empty as all the
129described parameters are shared. As no filename= option is given, fio
c2b1e753
JA
130makes up a filename for each of the jobs as it sees fit. On the command
131line, this job would look as follows:
132
133$ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
134
71bfa161
JA
135
136Lets look at an example that have a number of processes writing randomly
137to files.
138
139; -- start job file --
140[random-writers]
141ioengine=libaio
142iodepth=4
143rw=randwrite
144bs=32k
145direct=0
146size=64m
147numjobs=4
148
149; -- end job file --
150
151Here we have no global section, as we only have one job defined anyway.
152We want to use async io here, with a depth of 4 for each file. We also
153increased the buffer size used to 32KiB and define numjobs to 4 to
154fork 4 identical jobs. The result is 4 processes each randomly writing
b4692828
JA
155to their own 64MiB file. Instead of using the above job file, you could
156have given the parameters on the command line. For this case, you would
157specify:
158
159$ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
71bfa161
JA
160
161fio ships with a few example job files, you can also look there for
162inspiration.
163
164
1655.0 Detailed list of parameters
166-------------------------------
167
168This section describes in details each parameter associated with a job.
169Some parameters take an option of a given type, such as an integer or
170a string. The following types are used:
171
172str String. This is a sequence of alpha characters.
6d16ecb6 173int Integer. A whole number value, can be negative. If prefixed with
bf9a3edb 174 0x, the integer is assumed to be of base 16 (hexadecimal).
71bfa161
JA
175siint SI integer. A whole number value, which may contain a postfix
176 describing the base of the number. Accepted postfixes are k/m/g,
6c219763 177 meaning kilo, mega, and giga. So if you want to specify 4096,
71bfa161
JA
178 you could either write out '4096' or just give 4k. The postfixes
179 signify base 2 values, so 1024 is 1k and 1024k is 1m and so on.
43159d18 180 If the option accepts an upper and lower range, use a colon ':'
bf9a3edb 181 or minus '-' to separate such values. See irange.
71bfa161
JA
182bool Boolean. Usually parsed as an integer, however only defined for
183 true and false (1 and 0).
184irange Integer range with postfix. Allows value range to be given, such
bf9a3edb 185 as 1024-4096. A colon may also be used as the separator, eg
0c9baf91
JA
186 1k:4k. If the option allows two sets of ranges, they can be
187 specified with a ',' or '/' delimiter: 1k-4k/8k-32k. Also see
188 siint.
71bfa161
JA
189
190With the above in mind, here follows the complete list of fio job
191parameters.
192
193name=str ASCII name of the job. This may be used to override the
194 name printed by fio for this job. Otherwise the job
c2b1e753 195 name is used. On the command line this parameter has the
6c219763 196 special purpose of also signaling the start of a new
c2b1e753 197 job.
71bfa161 198
61697c37
JA
199description=str Text description of the job. Doesn't do anything except
200 dump this text description when this job is run. It's
201 not parsed.
202
71bfa161
JA
203directory=str Prefix filenames with this directory. Used to places files
204 in a different location than "./".
205
206filename=str Fio normally makes up a filename based on the job name,
207 thread number, and file number. If you want to share
208 files between threads in a job or several jobs, specify
ed92ac0c
JA
209 a filename for each of them to override the default. If
210 the ioengine used is 'net', the filename is the host and
9f9214f2 211 port to connect to in the format of =host/port. If the
af52b345 212 ioengine is file based, you can specify a number of files
bf9a3edb 213 by separating the names with a ':' colon. So if you wanted
af52b345 214 a job to open /dev/sda and /dev/sdb as the two working files,
66159828
JA
215 you would use filename=/dev/sda:/dev/sdb. '-' is a reserved
216 name, meaning stdin or stdout. Which of the two depends
217 on the read/write direction set.
71bfa161 218
bbf6b540
JA
219opendir=str Tell fio to recursively add any file it can find in this
220 directory and down the file system tree.
221
4d4e80f2
JA
222lockfile=str Fio defaults to not doing any locking files before it does
223 IO to them. If a file or file descriptor is shared, fio
224 can serialize IO to that file to make the end result
225 consistent. This is usual for emulating real workloads that
226 share files. The lock modes are:
227
228 none No locking. The default.
229 exclusive Only one thread/process may do IO,
230 excluding all others.
231 readwrite Read-write locking on the file. Many
232 readers may access the file at the
233 same time, but writes get exclusive
234 access.
235
236 The option may be post-fixed with a lock batch number. If
237 set, then each thread/process may do that amount of IOs to
bf9a3edb 238 the file before giving up the lock. Since lock acquisition is
4d4e80f2 239 expensive, batching the lock/unlocks will speed up IO.
29c1349f 240
d3aad8f2 241readwrite=str
71bfa161
JA
242rw=str Type of io pattern. Accepted values are:
243
244 read Sequential reads
245 write Sequential writes
246 randwrite Random writes
247 randread Random reads
248 rw Sequential mixed reads and writes
249 randrw Random mixed reads and writes
250
251 For the mixed io types, the default is to split them 50/50.
252 For certain types of io the result may still be skewed a bit,
211097b2
JA
253 since the speed may be different. It is possible to specify
254 a number of IO's to do before getting a new offset - this
255 is only useful for random IO, where fio would normally
256 generate a new random offset for every IO. If you append
257 eg 8 to randread, you would get a new random offset for
258 every 8 IO's. The result would be a seek for only every 8
259 IO's, instead of for every IO. Use rw=randread:8 to specify
260 that.
71bfa161 261
ee738499
JA
262randrepeat=bool For random IO workloads, seed the generator in a predictable
263 way so that results are repeatable across repetitions.
264
d2f3ac35
JA
265fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
266 on what IO patterns it is likely to issue. Sometimes you
267 want to test specific IO patterns without telling the
268 kernel about it, in which case you can disable this option.
269 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
270 IO and POSIX_FADV_RANDOM for random IO.
271
7616cafe
JA
272size=siint The total size of file io for this job. Fio will run until
273 this many bytes has been transferred, unless runtime is
274 limited by other options (such as 'runtime', for instance).
275 Unless specific nr_files and filesize options are given,
276 fio will divide this size between the available files
277 specified by the job.
71bfa161 278
9c60ce64
JA
279filesize=siint Individual file sizes. May be a range, in which case fio
280 will select sizes for files at random within the given range
281 and limited to 'size' in total (if that is given). If not
282 given, each created file is the same size.
283
aa31f1f1
SL
284fill_device=bool Sets size to something really large and waits for ENOSPC (no
285 space left on device) as the terminating condition. Only makes
286 sense with sequential write.
287
d3aad8f2 288blocksize=siint
f90eff5a
JA
289bs=siint The block size used for the io units. Defaults to 4k. Values
290 can be given for both read and writes. If a single siint is
291 given, it will apply to both. If a second siint is specified
292 after a comma, it will apply to writes only. In other words,
293 the format is either bs=read_and_write or bs=read,write.
294 bs=4k,8k will thus use 4k blocks for reads, and 8k blocks
787f7e95
JA
295 for writes. If you only wish to set the write size, you
296 can do so by passing an empty read size - bs=,8k will set
297 8k for writes and leave the read default value.
a00735e6 298
d3aad8f2 299blocksize_range=irange
71bfa161
JA
300bsrange=irange Instead of giving a single block size, specify a range
301 and fio will mix the issued io block sizes. The issued
302 io unit will always be a multiple of the minimum value
f90eff5a
JA
303 given (also see bs_unaligned). Applies to both reads and
304 writes, however a second range can be given after a comma.
305 See bs=.
a00735e6 306
564ca972
JA
307bssplit=str Sometimes you want even finer grained control of the
308 block sizes issued, not just an even split between them.
309 This option allows you to weight various block sizes,
310 so that you are able to define a specific amount of
311 block sizes issued. The format for this option is:
312
313 bssplit=blocksize/percentage:blocksize/percentage
314
315 for as many block sizes as needed. So if you want to define
316 a workload that has 50% 64k blocks, 10% 4k blocks, and
317 40% 32k blocks, you would write:
318
319 bssplit=4k/10:64k/50:32k/40
320
321 Ordering does not matter. If the percentage is left blank,
322 fio will fill in the remaining values evenly. So a bssplit
323 option like this one:
324
325 bssplit=4k/50:1k/:32k/
326
327 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
328 always add up to 100, if bssplit is given a range that adds
329 up to more, it will error out.
330
d3aad8f2 331blocksize_unaligned
690adba3
JA
332bs_unaligned If this option is given, any byte size value within bsrange
333 may be used as a block range. This typically wont work with
334 direct IO, as that normally requires sector alignment.
71bfa161 335
e9459e5a
JA
336zero_buffers If this option is given, fio will init the IO buffers to
337 all zeroes. The default is to fill them with random data.
338
5973cafb
JA
339refill_buffers If this option is given, fio will refill the IO buffers
340 on every submit. The default is to only fill it at init
341 time and reuse that data. Only makes sense if zero_buffers
41ccd845
JA
342 isn't specified, naturally. If data verification is enabled,
343 refill_buffers is also automatically enabled.
5973cafb 344
71bfa161
JA
345nrfiles=int Number of files to use for this job. Defaults to 1.
346
390b1537
JA
347openfiles=int Number of files to keep open at the same time. Defaults to
348 the same as nrfiles, can be set smaller to limit the number
349 simultaneous opens.
350
5af1c6f3
JA
351file_service_type=str Defines how fio decides which file from a job to
352 service next. The following types are defined:
353
354 random Just choose a file at random.
355
356 roundrobin Round robin over open files. This
357 is the default.
358
1907dbc6
JA
359 The string can have a number appended, indicating how
360 often to switch to a new file. So if option random:4 is
361 given, fio will switch to a new random file after 4 ios
362 have been issued.
363
71bfa161
JA
364ioengine=str Defines how the job issues io to the file. The following
365 types are defined:
366
367 sync Basic read(2) or write(2) io. lseek(2) is
368 used to position the io location.
369
a31041ea 370 psync Basic pread(2) or pwrite(2) io.
371
e05af9e5 372 vsync Basic readv(2) or writev(2) IO.
1d2af02a 373
71bfa161
JA
374 libaio Linux native asynchronous io.
375
376 posixaio glibc posix asynchronous io.
377
417f0068
JA
378 solarisaio Solaris native asynchronous io.
379
71bfa161
JA
380 mmap File is memory mapped and data copied
381 to/from using memcpy(3).
382
383 splice splice(2) is used to transfer the data and
384 vmsplice(2) to transfer data from user
385 space to the kernel.
386
d0ff85df
JA
387 syslet-rw Use the syslet system calls to make
388 regular read/write async.
389
71bfa161 390 sg SCSI generic sg v3 io. May either be
6c219763 391 synchronous using the SG_IO ioctl, or if
71bfa161
JA
392 the target is an sg character device
393 we use read(2) and write(2) for asynchronous
394 io.
395
a94ea28b
JA
396 null Doesn't transfer any data, just pretends
397 to. This is mainly used to exercise fio
398 itself and for debugging/testing purposes.
399
ed92ac0c
JA
400 net Transfer over the network to given host:port.
401 'filename' must be set appropriately to
9f9214f2 402 filename=host/port regardless of send
ed92ac0c
JA
403 or receive, if the latter only the port
404 argument is used.
405
9cce02e8
JA
406 netsplice Like net, but uses splice/vmsplice to
407 map data and send/receive.
408
53aec0a4 409 cpuio Doesn't transfer any data, but burns CPU
ba0fbe10
JA
410 cycles according to the cpuload= and
411 cpucycle= options. Setting cpuload=85
412 will cause that job to do nothing but burn
36ecec83
GP
413 85% of the CPU. In case of SMP machines,
414 use numjobs=<no_of_cpu> to get desired CPU
415 usage, as the cpuload only loads a single
416 CPU at the desired rate.
ba0fbe10 417
e9a1806f
JA
418 guasi The GUASI IO engine is the Generic Userspace
419 Asyncronous Syscall Interface approach
420 to async IO. See
421
422 http://www.xmailserver.org/guasi-lib.html
423
424 for more info on GUASI.
425
8a7bd877
JA
426 external Prefix to specify loading an external
427 IO engine object file. Append the engine
428 filename, eg ioengine=external:/tmp/foo.o
429 to load ioengine foo.o in /tmp.
430
71bfa161
JA
431iodepth=int This defines how many io units to keep in flight against
432 the file. The default is 1 for each file defined in this
433 job, can be overridden with a larger value for higher
434 concurrency.
435
4950421a 436iodepth_batch_submit=int
cb5ab512 437iodepth_batch=int This defines how many pieces of IO to submit at once.
89e820f6
JA
438 It defaults to 1 which means that we submit each IO
439 as soon as it is available, but can be raised to submit
440 bigger batches of IO at the time.
cb5ab512 441
4950421a
JA
442iodepth_batch_complete=int This defines how many pieces of IO to retrieve
443 at once. It defaults to 1 which means that we'll ask
444 for a minimum of 1 IO in the retrieval process from
445 the kernel. The IO retrieval will go on until we
446 hit the limit set by iodepth_low. If this variable is
447 set to 0, then fio will always check for completed
448 events before queuing more IO. This helps reduce
449 IO latency, at the cost of more retrieval system calls.
450
e916b390
JA
451iodepth_low=int The low water mark indicating when to start filling
452 the queue again. Defaults to the same as iodepth, meaning
453 that fio will attempt to keep the queue full at all times.
454 If iodepth is set to eg 16 and iodepth_low is set to 4, then
455 after fio has filled the queue of 16 requests, it will let
456 the depth drain down to 4 before starting to fill it again.
457
71bfa161 458direct=bool If value is true, use non-buffered io. This is usually
76a43db4
JA
459 O_DIRECT.
460
461buffered=bool If value is true, use buffered io. This is the opposite
462 of the 'direct' option. Defaults to true.
71bfa161
JA
463
464offset=siint Start io at the given offset in the file. The data before
465 the given offset will not be touched. This effectively
466 caps the file size at real_size - offset.
467
468fsync=int If writing to a file, issue a sync of the dirty data
469 for every number of blocks given. For example, if you give
470 32 as a parameter, fio will sync the file for every 32
471 writes issued. If fio is using non-buffered io, we may
472 not sync the file. The exception is the sg io engine, which
6c219763 473 synchronizes the disk cache anyway.
71bfa161 474
5036fc1e
JA
475overwrite=bool If true, writes to a file will always overwrite existing
476 data. If the file doesn't already exist, it will be
477 created before the write phase begins. If the file exists
478 and is large enough for the specified write phase, nothing
479 will be done.
71bfa161
JA
480
481end_fsync=bool If true, fsync file contents when the job exits.
482
ebb1415f
JA
483fsync_on_close=bool If true, fio will fsync() a dirty file on close.
484 This differs from end_fsync in that it will happen on every
485 file close, not just at the end of the job.
486
71bfa161
JA
487rwmixread=int How large a percentage of the mix should be reads.
488
489rwmixwrite=int How large a percentage of the mix should be writes. If both
490 rwmixread and rwmixwrite is given and the values do not add
491 up to 100%, the latter of the two will be used to override
492 the first.
493
bb8895e0
JA
494norandommap Normally fio will cover every block of the file when doing
495 random IO. If this option is given, fio will just get a
496 new random offset without looking at past io history. This
497 means that some blocks may not be read or written, and that
498 some blocks may be read/written more than once. This option
7616cafe
JA
499 is mutually exclusive with verify= for that reason, since
500 fio doesn't track potential block rewrites which may alter
501 the calculated checksum for that block.
bb8895e0 502
2b386d25
JA
503softrandommap See norandommap. If fio runs with the random block map enabled
504 and it fails to allocate the map, if this option is set it
505 will continue without a random block map. As coverage will
506 not be as complete as with random maps, this option is
507 disabled by default.
508
71bfa161
JA
509nice=int Run the job with the given nice value. See man nice(2).
510
511prio=int Set the io priority value of this job. Linux limits us to
512 a positive value between 0 and 7, with 0 being the highest.
513 See man ionice(1).
514
515prioclass=int Set the io priority class. See man ionice(1).
516
517thinktime=int Stall the job x microseconds after an io has completed before
518 issuing the next. May be used to simulate processing being
48097d5c
JA
519 done by an application. See thinktime_blocks and
520 thinktime_spin.
521
522thinktime_spin=int
523 Only valid if thinktime is set - pretend to spend CPU time
524 doing something with the data received, before falling back
525 to sleeping for the rest of the period specified by
526 thinktime.
9c1f7434
JA
527
528thinktime_blocks
529 Only valid if thinktime is set - control how many blocks
530 to issue, before waiting 'thinktime' usecs. If not set,
531 defaults to 1 which will make fio wait 'thinktime' usecs
532 after every block.
71bfa161
JA
533
534rate=int Cap the bandwidth used by this job to this number of KiB/sec.
535
536ratemin=int Tell fio to do whatever it can to maintain at least this
4e991c23
JA
537 bandwidth. Failing to meet this requirement, will cause
538 the job to exit.
539
540rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
541 as rate, just specified independently of bandwidth. If the
542 job is given a block size range instead of a fixed value,
543 the smallest block size is used as the metric.
544
545rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
546 the job to exit.
71bfa161
JA
547
548ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
6c219763 549 of milliseconds.
71bfa161
JA
550
551cpumask=int Set the CPU affinity of this job. The parameter given is a
a08bc17f
JA
552 bitmask of allowed CPU's the job may run on. So if you want
553 the allowed CPUs to be 1 and 5, you would pass the decimal
554 value of (1 << 1 | 1 << 5), or 34. See man
7dbb6eba
JA
555 sched_setaffinity(2). This may not work on all supported
556 operating systems or kernel versions.
71bfa161 557
d2e268b0
JA
558cpus_allowed=str Controls the same options as cpumask, but it allows a text
559 setting of the permitted CPUs instead. So to use CPUs 1 and
560 5, you would specify cpus_allowed=1,5.
561
71bfa161
JA
562startdelay=int Start this job the specified number of seconds after fio
563 has started. Only useful if the job file contains several
564 jobs, and you want to delay starting some jobs to a certain
565 time.
566
03b74b3e 567runtime=int Tell fio to terminate processing after the specified number
71bfa161
JA
568 of seconds. It can be quite hard to determine for how long
569 a specified job will run, so this parameter is handy to
570 cap the total runtime to a given time.
571
cf4464ca 572time_based If set, fio will run for the duration of the runtime
bf9a3edb 573 specified even if the file(s) are completely read or
cf4464ca
JA
574 written. It will simply loop over the same workload
575 as many times as the runtime allows.
576
71bfa161
JA
577invalidate=bool Invalidate the buffer/page cache parts for this file prior
578 to starting io. Defaults to true.
579
580sync=bool Use sync io for buffered writes. For the majority of the
581 io engines, this means using O_SYNC.
582
d3aad8f2 583iomem=str
71bfa161
JA
584mem=str Fio can use various types of memory as the io unit buffer.
585 The allowed values are:
586
587 malloc Use memory from malloc(3) as the buffers.
588
589 shm Use shared memory as the buffers. Allocated
590 through shmget(2).
591
74b025b0
JA
592 shmhuge Same as shm, but use huge pages as backing.
593
313cb206
JA
594 mmap Use mmap to allocate buffers. May either be
595 anonymous memory, or can be file backed if
596 a filename is given after the option. The
597 format is mem=mmap:/path/to/file.
71bfa161 598
d0bdaf49
JA
599 mmaphuge Use a memory mapped huge file as the buffer
600 backing. Append filename after mmaphuge, ala
601 mem=mmaphuge:/hugetlbfs/file
602
71bfa161 603 The area allocated is a function of the maximum allowed
5394ae5f
JA
604 bs size for the job, multiplied by the io depth given. Note
605 that for shmhuge and mmaphuge to work, the system must have
606 free huge pages allocated. This can normally be checked
607 and set by reading/writing /proc/sys/vm/nr_hugepages on a
608 Linux system. Fio assumes a huge page is 4MiB in size. So
609 to calculate the number of huge pages you need for a given
610 job file, add up the io depth of all jobs (normally one unless
611 iodepth= is used) and multiply by the maximum bs set. Then
612 divide that number by the huge page size. You can see the
613 size of the huge pages in /proc/meminfo. If no huge pages
614 are allocated by having a non-zero number in nr_hugepages,
56bb17f2 615 using mmaphuge or shmhuge will fail. Also see hugepage-size.
5394ae5f
JA
616
617 mmaphuge also needs to have hugetlbfs mounted and the file
618 location should point there. So if it's mounted in /huge,
619 you would use mem=mmaphuge:/huge/somefile.
71bfa161 620
56bb17f2
JA
621hugepage-size=siint
622 Defines the size of a huge page. Must at least be equal
623 to the system setting, see /proc/meminfo. Defaults to 4MiB.
c51074e7
JA
624 Should probably always be a multiple of megabytes, so using
625 hugepage-size=Xm is the preferred way to set this to avoid
626 setting a non-pow-2 bad value.
56bb17f2 627
71bfa161
JA
628exitall When one job finishes, terminate the rest. The default is
629 to wait for each job to finish, sometimes that is not the
630 desired action.
631
632bwavgtime=int Average the calculated bandwidth over the given time. Value
6c219763 633 is specified in milliseconds.
71bfa161
JA
634
635create_serialize=bool If true, serialize the file creating for the jobs.
636 This may be handy to avoid interleaving of data
637 files, which may greatly depend on the filesystem
638 used and even the number of processors in the system.
639
640create_fsync=bool fsync the data file after creation. This is the
641 default.
642
e545a6ce 643unlink=bool Unlink the job files when done. Not the default, as repeated
bf9a3edb
JA
644 runs of that job would then waste time recreating the file
645 set again and again.
71bfa161
JA
646
647loops=int Run the specified number of iterations of this job. Used
648 to repeat the same workload a given number of times. Defaults
649 to 1.
650
68e1f29a 651do_verify=bool Run the verify phase after a write phase. Only makes sense if
e84c73a8
SL
652 verify is set. Defaults to 1.
653
71bfa161
JA
654verify=str If writing to a file, fio can verify the file contents
655 after each iteration of the job. The allowed values are:
656
657 md5 Use an md5 sum of the data area and store
658 it in the header of each block.
659
17dc34df
JA
660 crc64 Use an experimental crc64 sum of the data
661 area and store it in the header of each
662 block.
663
bac39e0e
JA
664 crc32c Use a crc32c sum of the data area and store
665 it in the header of each block.
666
3845591f
JA
667 crc32c-intel Use hardware assisted crc32c calcuation
668 provided on SSE4.2 enabled processors.
669
71bfa161
JA
670 crc32 Use a crc32 sum of the data area and store
671 it in the header of each block.
672
969f7ed3
JA
673 crc16 Use a crc16 sum of the data area and store
674 it in the header of each block.
675
17dc34df
JA
676 crc7 Use a crc7 sum of the data area and store
677 it in the header of each block.
678
cd14cc10
JA
679 sha512 Use sha512 as the checksum function.
680
681 sha256 Use sha256 as the checksum function.
682
7437ee87
SL
683 meta Write extra information about each io
684 (timestamp, block number etc.). The block
685 number is verified.
686
36690c9b
JA
687 null Only pretend to verify. Useful for testing
688 internals with ioengine=null, not for much
689 else.
690
6c219763 691 This option can be used for repeated burn-in tests of a
71bfa161
JA
692 system to make sure that the written data is also
693 correctly read back.
694
160b966d
JA
695verifysort=bool If set, fio will sort written verify blocks when it deems
696 it faster to read them back in a sorted manner. This is
697 often the case when overwriting an existing file, since
698 the blocks are already laid out in the file system. You
699 can ignore this option unless doing huge amounts of really
700 fast IO where the red-black tree sorting CPU time becomes
701 significant.
3f9f4e26 702
a59e170d 703verify_offset=siint Swap the verification header with data somewhere else
546a9142
SL
704 in the block before writing. Its swapped back before
705 verifying.
706
a59e170d 707verify_interval=siint Write the verification header at a finer granularity
3f9f4e26
SL
708 than the blocksize. It will be written for chunks the
709 size of header_interval. blocksize should divide this
710 evenly.
90059d65 711
e28218f3
SL
712verify_pattern=int If set, fio will fill the io buffers with this
713 pattern. Fio defaults to filling with totally random
714 bytes, but sometimes it's interesting to fill with a known
715 pattern for io verification purposes. Depending on the
716 width of the pattern, fio will fill 1/2/3/4 bytes of the
717 buffer at the time. The verify_pattern cannot be larger than
718 a 32-bit quantity.
719
68e1f29a 720verify_fatal=bool Normally fio will keep checking the entire contents
a12a3b4d
JA
721 before quitting on a block verification failure. If this
722 option is set, fio will exit the job on the first observed
723 failure.
160b966d 724
71bfa161
JA
725stonewall Wait for preceeding jobs in the job file to exit, before
726 starting this one. Can be used to insert serialization
b3d62a75
JA
727 points in the job file. A stone wall also implies starting
728 a new reporting group.
729
730new_group Start a new reporting group. If this option isn't given,
731 jobs in a file will be part of the same reporting group
bf9a3edb 732 unless separated by a stone wall (or if it's a group
b3d62a75 733 by itself, with the numjobs option).
71bfa161
JA
734
735numjobs=int Create the specified number of clones of this job. May be
736 used to setup a larger number of threads/processes doing
fa28c85a
JA
737 the same thing. We regard that grouping of jobs as a
738 specific group.
739
740group_reporting If 'numjobs' is set, it may be interesting to display
741 statistics for the group as a whole instead of for each
742 individual job. This is especially true of 'numjobs' is
743 large, looking at individual thread/process output quickly
744 becomes unwieldy. If 'group_reporting' is specified, fio
745 will show the final report per-group instead of per-job.
71bfa161
JA
746
747thread fio defaults to forking jobs, however if this option is
748 given, fio will use pthread_create(3) to create threads
749 instead.
750
751zonesize=siint Divide a file into zones of the specified size. See zoneskip.
752
753zoneskip=siint Skip the specified number of bytes when zonesize data has
754 been read. The two zone options can be used to only do
755 io on zones of a file.
756
076efc7c
JA
757write_iolog=str Write the issued io patterns to the specified file. See
758 read_iolog.
71bfa161 759
076efc7c 760read_iolog=str Open an iolog with the specified file name and replay the
71bfa161 761 io patterns it contains. This can be used to store a
6df8adaa
JA
762 workload and replay it sometime later. The iolog given
763 may also be a blktrace binary file, which allows fio
764 to replay a workload captured by blktrace. See blktrace
765 for how to capture such logging data. For blktrace replay,
766 the file needs to be turned into a blkparse binary data
767 file first (blktrace <device> -d file_for_fio.bin).
71bfa161
JA
768
769write_bw_log If given, write a bandwidth log of the jobs in this job
770 file. Can be used to store data of the bandwidth of the
e0da9bc2
JA
771 jobs in their lifetime. The included fio_generate_plots
772 script uses gnuplot to turn these text files into nice
773 graphs.
71bfa161
JA
774
775write_lat_log Same as write_bw_log, except that this option stores io
776 completion latencies instead.
777
778lockmem=siint Pin down the specified amount of memory with mlock(2). Can
779 potentially be used instead of removing memory or booting
780 with less memory to simulate a smaller amount of memory.
781
782exec_prerun=str Before running this job, issue the command specified
783 through system(3).
784
785exec_postrun=str After the job completes, issue the command specified
786 though system(3).
787
788ioscheduler=str Attempt to switch the device hosting the file to the specified
789 io scheduler before running.
790
791cpuload=int If the job is a CPU cycle eater, attempt to use the specified
792 percentage of CPU cycles.
793
794cpuchunks=int If the job is a CPU cycle eater, split the load into
6c219763 795 cycles of the given time. In milliseconds.
71bfa161 796
0a839f30
JA
797disk_util=bool Generate disk utilization statistics, if the platform
798 supports it. Defaults to on.
799
71bfa161
JA
800
8016.0 Interpreting the output
802---------------------------
803
804fio spits out a lot of output. While running, fio will display the
805status of the jobs created. An example of that would be:
806
73c8b082 807Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
71bfa161
JA
808
809The characters inside the square brackets denote the current status of
810each thread. The possible values (in typical life cycle order) are:
811
812Idle Run
813---- ---
814P Thread setup, but not started.
815C Thread created.
816I Thread initialized, waiting.
817 R Running, doing sequential reads.
818 r Running, doing random reads.
819 W Running, doing sequential writes.
820 w Running, doing random writes.
821 M Running, doing mixed sequential reads/writes.
822 m Running, doing mixed random reads/writes.
823 F Running, currently waiting for fsync()
824V Running, doing verification of written data.
825E Thread exited, not reaped by main thread yet.
826_ Thread reaped.
827
828The other values are fairly self explanatory - number of threads
c9f60304
JA
829currently running and doing io, rate of io since last check (read speed
830listed first, then write speed), and the estimated completion percentage
831and time for the running group. It's impossible to estimate runtime of
832the following groups (if any).
71bfa161
JA
833
834When fio is done (or interrupted by ctrl-c), it will show the data for
835each thread, group of threads, and disks in that order. For each data
836direction, the output looks like:
837
838Client1 (g=0): err= 0:
839 write: io= 32MiB, bw= 666KiB/s, runt= 50320msec
6104ddb6
JA
840 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
841 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
842 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
e7823a94 843 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
71619dc2 844 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
838bc709
JA
845 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
846 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
30061b97 847 issued r/w: total=0/32768, short=0/0
8abdce66
JA
848 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
849 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
71bfa161
JA
850
851The client number is printed, along with the group id and error of that
852thread. Below is the io statistics, here for writes. In the order listed,
853they denote:
854
855io= Number of megabytes io performed
856bw= Average bandwidth rate
857runt= The runtime of that thread
72fbda2a 858 slat= Submission latency (avg being the average, stdev being the
71bfa161
JA
859 standard deviation). This is the time it took to submit
860 the io. For sync io, the slat is really the completion
8a35c71e 861 latency, since queue/complete is one operation there. This
bf9a3edb 862 value can be in milliseconds or microseconds, fio will choose
8a35c71e 863 the most appropriate base and print that. In the example
bf9a3edb 864 above, milliseconds is the best scale.
71bfa161
JA
865 clat= Completion latency. Same names as slat, this denotes the
866 time from submission to completion of the io pieces. For
867 sync io, clat will usually be equal (or very close) to 0,
868 as the time from submit to complete is basically just
869 CPU time (io has already been done, see slat explanation).
870 bw= Bandwidth. Same names as the xlat stats, but also includes
871 an approximate percentage of total aggregate bandwidth
872 this thread received in this group. This last value is
873 only really useful if the threads in this group are on the
874 same disk, since they are then competing for disk access.
875cpu= CPU usage. User and system time, along with the number
e7823a94
JA
876 of context switches this thread went through, usage of
877 system and user time, and finally the number of major
878 and minor page faults.
71619dc2
JA
879IO depths= The distribution of io depths over the job life time. The
880 numbers are divided into powers of 2, so for example the
881 16= entries includes depths up to that value but higher
882 than the previous entry. In other words, it covers the
883 range from 16 to 31.
838bc709
JA
884IO submit= How many pieces of IO were submitting in a single submit
885 call. Each entry denotes that amount and below, until
886 the previous entry - eg, 8=100% mean that we submitted
887 anywhere in between 5-8 ios per submit call.
888IO complete= Like the above submit number, but for completions instead.
30061b97
JA
889IO issued= The number of read/write requests issued, and how many
890 of them were short.
ec118304
JA
891IO latencies= The distribution of IO completion latencies. This is the
892 time from when IO leaves fio and when it gets completed.
893 The numbers follow the same pattern as the IO depths,
894 meaning that 2=1.6% means that 1.6% of the IO completed
8abdce66
JA
895 within 2 msecs, 20=12.8% means that 12.8% of the IO
896 took more than 10 msecs, but less than (or equal to) 20 msecs.
71bfa161
JA
897
898After each client has been listed, the group statistics are printed. They
899will look like this:
900
901Run status group 0 (all jobs):
902 READ: io=64MiB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
903 WRITE: io=64MiB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
904
905For each data direction, it prints:
906
907io= Number of megabytes io performed.
908aggrb= Aggregate bandwidth of threads in this group.
909minb= The minimum average bandwidth a thread saw.
910maxb= The maximum average bandwidth a thread saw.
911mint= The smallest runtime of the threads in that group.
912maxt= The longest runtime of the threads in that group.
913
914And finally, the disk statistics are printed. They will look like this:
915
916Disk stats (read/write):
917 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
918
919Each value is printed for both reads and writes, with reads first. The
920numbers denote:
921
922ios= Number of ios performed by all groups.
923merge= Number of merges io the io scheduler.
924ticks= Number of ticks we kept the disk busy.
925io_queue= Total time spent in the disk queue.
926util= The disk utilization. A value of 100% means we kept the disk
927 busy constantly, 50% would be a disk idling half of the time.
928
929
9307.0 Terse output
931----------------
932
933For scripted usage where you typically want to generate tables or graphs
6af019c9 934of the results, fio can output the results in a semicolon separated format.
71bfa161
JA
935The format is one long line of values, such as:
936
6af019c9
JA
937client1;0;0;1906777;1090804;1790;0;0;0.000000;0.000000;0;0;0.000000;0.000000;929380;1152890;25.510151%;1078276.333333;128948.113404;0;0;0;0;0;0.000000;0.000000;0;0;0.000000;0.000000;0;0;0.000000%;0.000000;0.000000;100.000000%;0.000000%;324;100.0%;0.0%;0.0%;0.0%;0.0%;0.0%;0.0%;100.0%;0.0%;0.0%;0.0%;0.0%;0.0%
938;0.0%;0.0%;0.0%;0.0%;0.0%
71bfa161
JA
939
940Split up, the format is as follows:
941
942 jobname, groupid, error
943 READ status:
944 KiB IO, bandwidth (KiB/sec), runtime (msec)
945 Submission latency: min, max, mean, deviation
946 Completion latency: min, max, mean, deviation
6c219763 947 Bw: min, max, aggregate percentage of total, mean, deviation
71bfa161
JA
948 WRITE status:
949 KiB IO, bandwidth (KiB/sec), runtime (msec)
950 Submission latency: min, max, mean, deviation
951 Completion latency: min, max, mean, deviation
6c219763 952 Bw: min, max, aggregate percentage of total, mean, deviation
046ee302 953 CPU usage: user, system, context switches, major faults, minor faults
2270890c
JA
954 IO depths: <=1, 2, 4, 8, 16, 32, >=64
955 IO latencies: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, >=2000
956 Text description
71bfa161 957