Fix man page documentation for verify_pattern
[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 113
3c54bc46 114So let's look at a really simple job file that defines two processes, each
b22989b9 115randomly reading from a 128MB file.
71bfa161
JA
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 135
3c54bc46 136Let's look at an example that has a number of processes writing randomly
71bfa161
JA
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
b22989b9 153increased the buffer size used to 32KB and define numjobs to 4 to
71bfa161 154fork 4 identical jobs. The result is 4 processes each randomly writing
b22989b9 155to their own 64MB file. Instead of using the above job file, you could
b4692828
JA
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 160
74929ac2
JA
1614.1 Environment variables
162-------------------------
163
3c54bc46
AC
164fio also supports environment variable expansion in job files. Any
165substring of the form "${VARNAME}" as part of an option value (in other
166words, on the right of the `='), will be expanded to the value of the
167environment variable called VARNAME. If no such environment variable
168is defined, or VARNAME is the empty string, the empty string will be
169substituted.
170
171As an example, let's look at a sample fio invocation and job file:
172
173$ SIZE=64m NUMJOBS=4 fio jobfile.fio
174
175; -- start job file --
176[random-writers]
177rw=randwrite
178size=${SIZE}
179numjobs=${NUMJOBS}
180; -- end job file --
181
182This will expand to the following equivalent job file at runtime:
183
184; -- start job file --
185[random-writers]
186rw=randwrite
187size=64m
188numjobs=4
189; -- end job file --
190
71bfa161
JA
191fio ships with a few example job files, you can also look there for
192inspiration.
193
74929ac2
JA
1944.2 Reserved keywords
195---------------------
196
197Additionally, fio has a set of reserved keywords that will be replaced
198internally with the appropriate value. Those keywords are:
199
200$pagesize The architecture page size of the running system
201$mb_memory Megabytes of total memory in the system
202$ncpus Number of online available CPUs
203
204These can be used on the command line or in the job file, and will be
205automatically substituted with the current system values when the job
892a6ffc
JA
206is run. Simple math is also supported on these keywords, so you can
207perform actions like:
208
209size=8*$mb_memory
210
211and get that properly expanded to 8 times the size of memory in the
212machine.
74929ac2 213
71bfa161
JA
214
2155.0 Detailed list of parameters
216-------------------------------
217
218This section describes in details each parameter associated with a job.
219Some parameters take an option of a given type, such as an integer or
220a string. The following types are used:
221
222str String. This is a sequence of alpha characters.
b09da8fa 223time Integer with possible time suffix. In seconds unless otherwise
e417fd66
JA
224 specified, use eg 10m for 10 minutes. Accepts s/m/h for seconds,
225 minutes, and hours.
b09da8fa
JA
226int SI integer. A whole number value, which may contain a suffix
227 describing the base of the number. Accepted suffixes are k/m/g/t/p,
228 meaning kilo, mega, giga, tera, and peta. The suffix is not case
57fc29fa
JA
229 sensitive, and you may also include trailing 'b' (eg 'kb' is the same
230 as 'k'). So if you want to specify 4096, you could either write
b09da8fa 231 out '4096' or just give 4k. The suffixes signify base 2 values, so
57fc29fa
JA
232 1024 is 1k and 1024k is 1m and so on, unless the suffix is explicitly
233 set to a base 10 value using 'kib', 'mib', 'gib', etc. If that is the
234 case, then 1000 is used as the multiplier. This can be handy for
235 disks, since manufacturers generally use base 10 values when listing
236 the capacity of a drive. If the option accepts an upper and lower
237 range, use a colon ':' or minus '-' to separate such values. May also
238 include a prefix to indicate numbers base. If 0x is used, the number
239 is assumed to be hexadecimal. See irange.
71bfa161
JA
240bool Boolean. Usually parsed as an integer, however only defined for
241 true and false (1 and 0).
b09da8fa 242irange Integer range with suffix. Allows value range to be given, such
bf9a3edb 243 as 1024-4096. A colon may also be used as the separator, eg
0c9baf91
JA
244 1k:4k. If the option allows two sets of ranges, they can be
245 specified with a ',' or '/' delimiter: 1k-4k/8k-32k. Also see
f7fa2653 246 int.
71bfa161
JA
247
248With the above in mind, here follows the complete list of fio job
249parameters.
250
251name=str ASCII name of the job. This may be used to override the
252 name printed by fio for this job. Otherwise the job
c2b1e753 253 name is used. On the command line this parameter has the
6c219763 254 special purpose of also signaling the start of a new
c2b1e753 255 job.
71bfa161 256
61697c37
JA
257description=str Text description of the job. Doesn't do anything except
258 dump this text description when this job is run. It's
259 not parsed.
260
3776041e 261directory=str Prefix filenames with this directory. Used to place files
71bfa161
JA
262 in a different location than "./".
263
264filename=str Fio normally makes up a filename based on the job name,
265 thread number, and file number. If you want to share
266 files between threads in a job or several jobs, specify
ed92ac0c 267 a filename for each of them to override the default. If
414c2a3e
JA
268 the ioengine used is 'net', the filename is the host, port,
269 and protocol to use in the format of =host/port/protocol.
270 See ioengine=net for more. If the ioengine is file based, you
271 can specify a number of files by separating the names with a
272 ':' colon. So if you wanted a job to open /dev/sda and /dev/sdb
273 as the two working files, you would use
8e827d35
JA
274 filename=/dev/sda:/dev/sdb. If the wanted filename does need to
275 include a colon, then escape that with a '\' character. For
276 instance, if the filename is "/dev/dsk/foo@3,0:c", then you would
277 use filename="/dev/dsk/foo@3,0\:c". '-' is a reserved name,
278 meaning stdin or stdout. Which of the two depends on the read/write
414c2a3e 279 direction set.
71bfa161 280
bbf6b540
JA
281opendir=str Tell fio to recursively add any file it can find in this
282 directory and down the file system tree.
283
3776041e 284lockfile=str Fio defaults to not locking any files before it does
4d4e80f2
JA
285 IO to them. If a file or file descriptor is shared, fio
286 can serialize IO to that file to make the end result
287 consistent. This is usual for emulating real workloads that
288 share files. The lock modes are:
289
290 none No locking. The default.
291 exclusive Only one thread/process may do IO,
292 excluding all others.
293 readwrite Read-write locking on the file. Many
294 readers may access the file at the
295 same time, but writes get exclusive
296 access.
297
298 The option may be post-fixed with a lock batch number. If
299 set, then each thread/process may do that amount of IOs to
bf9a3edb 300 the file before giving up the lock. Since lock acquisition is
4d4e80f2 301 expensive, batching the lock/unlocks will speed up IO.
29c1349f 302
d3aad8f2 303readwrite=str
71bfa161
JA
304rw=str Type of io pattern. Accepted values are:
305
306 read Sequential reads
307 write Sequential writes
308 randwrite Random writes
309 randread Random reads
310 rw Sequential mixed reads and writes
311 randrw Random mixed reads and writes
312
313 For the mixed io types, the default is to split them 50/50.
314 For certain types of io the result may still be skewed a bit,
211097b2
JA
315 since the speed may be different. It is possible to specify
316 a number of IO's to do before getting a new offset - this
317 is only useful for random IO, where fio would normally
318 generate a new random offset for every IO. If you append
319 eg 8 to randread, you would get a new random offset for
320 every 8 IO's. The result would be a seek for only every 8
321 IO's, instead of for every IO. Use rw=randread:8 to specify
322 that.
71bfa161 323
90fef2d1
JA
324kb_base=int The base unit for a kilobyte. The defacto base is 2^10, 1024.
325 Storage manufacturers like to use 10^3 or 1000 as a base
326 ten unit instead, for obvious reasons. Allow values are
327 1024 or 1000, with 1024 being the default.
328
ee738499
JA
329randrepeat=bool For random IO workloads, seed the generator in a predictable
330 way so that results are repeatable across repetitions.
331
7bc8c2cf
JA
332fallocate=bool By default, fio will use fallocate() to advise the system
333 of the size of the file we are going to write. This can be
334 turned off with fallocate=0. May not be available on all
335 supported platforms.
336
d2f3ac35
JA
337fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
338 on what IO patterns it is likely to issue. Sometimes you
339 want to test specific IO patterns without telling the
340 kernel about it, in which case you can disable this option.
341 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
342 IO and POSIX_FADV_RANDOM for random IO.
343
f7fa2653 344size=int The total size of file io for this job. Fio will run until
7616cafe
JA
345 this many bytes has been transferred, unless runtime is
346 limited by other options (such as 'runtime', for instance).
3776041e 347 Unless specific nrfiles and filesize options are given,
7616cafe
JA
348 fio will divide this size between the available files
349 specified by the job.
71bfa161 350
f7fa2653 351filesize=int Individual file sizes. May be a range, in which case fio
9c60ce64
JA
352 will select sizes for files at random within the given range
353 and limited to 'size' in total (if that is given). If not
354 given, each created file is the same size.
355
aa31f1f1
SL
356fill_device=bool Sets size to something really large and waits for ENOSPC (no
357 space left on device) as the terminating condition. Only makes
3ce9dcaf
JA
358 sense with sequential write. For a read workload, the mount
359 point will be filled first then IO started on the result.
aa31f1f1 360
f7fa2653
JA
361blocksize=int
362bs=int The block size used for the io units. Defaults to 4k. Values
363 can be given for both read and writes. If a single int is
364 given, it will apply to both. If a second int is specified
f90eff5a
JA
365 after a comma, it will apply to writes only. In other words,
366 the format is either bs=read_and_write or bs=read,write.
367 bs=4k,8k will thus use 4k blocks for reads, and 8k blocks
787f7e95
JA
368 for writes. If you only wish to set the write size, you
369 can do so by passing an empty read size - bs=,8k will set
370 8k for writes and leave the read default value.
a00735e6 371
2b7a01d0
JA
372blockalign=int
373ba=int At what boundary to align random IO offsets. Defaults to
374 the same as 'blocksize' the minimum blocksize given.
375 Minimum alignment is typically 512b for using direct IO,
376 though it usually depends on the hardware block size. This
377 option is mutually exclusive with using a random map for
378 files, so it will turn off that option.
379
d3aad8f2 380blocksize_range=irange
71bfa161
JA
381bsrange=irange Instead of giving a single block size, specify a range
382 and fio will mix the issued io block sizes. The issued
383 io unit will always be a multiple of the minimum value
f90eff5a
JA
384 given (also see bs_unaligned). Applies to both reads and
385 writes, however a second range can be given after a comma.
386 See bs=.
a00735e6 387
564ca972
JA
388bssplit=str Sometimes you want even finer grained control of the
389 block sizes issued, not just an even split between them.
390 This option allows you to weight various block sizes,
391 so that you are able to define a specific amount of
392 block sizes issued. The format for this option is:
393
394 bssplit=blocksize/percentage:blocksize/percentage
395
396 for as many block sizes as needed. So if you want to define
397 a workload that has 50% 64k blocks, 10% 4k blocks, and
398 40% 32k blocks, you would write:
399
400 bssplit=4k/10:64k/50:32k/40
401
402 Ordering does not matter. If the percentage is left blank,
403 fio will fill in the remaining values evenly. So a bssplit
404 option like this one:
405
406 bssplit=4k/50:1k/:32k/
407
408 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
409 always add up to 100, if bssplit is given a range that adds
410 up to more, it will error out.
411
720e84ad
JA
412 bssplit also supports giving separate splits to reads and
413 writes. The format is identical to what bs= accepts. You
414 have to separate the read and write parts with a comma. So
415 if you want a workload that has 50% 2k reads and 50% 4k reads,
416 while having 90% 4k writes and 10% 8k writes, you would
417 specify:
418
419 bssplit=2k/50:4k/50,4k/90,8k/10
420
d3aad8f2 421blocksize_unaligned
690adba3
JA
422bs_unaligned If this option is given, any byte size value within bsrange
423 may be used as a block range. This typically wont work with
424 direct IO, as that normally requires sector alignment.
71bfa161 425
e9459e5a
JA
426zero_buffers If this option is given, fio will init the IO buffers to
427 all zeroes. The default is to fill them with random data.
428
5973cafb
JA
429refill_buffers If this option is given, fio will refill the IO buffers
430 on every submit. The default is to only fill it at init
431 time and reuse that data. Only makes sense if zero_buffers
41ccd845
JA
432 isn't specified, naturally. If data verification is enabled,
433 refill_buffers is also automatically enabled.
5973cafb 434
71bfa161
JA
435nrfiles=int Number of files to use for this job. Defaults to 1.
436
390b1537
JA
437openfiles=int Number of files to keep open at the same time. Defaults to
438 the same as nrfiles, can be set smaller to limit the number
439 simultaneous opens.
440
5af1c6f3
JA
441file_service_type=str Defines how fio decides which file from a job to
442 service next. The following types are defined:
443
444 random Just choose a file at random.
445
446 roundrobin Round robin over open files. This
447 is the default.
448
a086c257
JA
449 sequential Finish one file before moving on to
450 the next. Multiple files can still be
451 open depending on 'openfiles'.
452
1907dbc6
JA
453 The string can have a number appended, indicating how
454 often to switch to a new file. So if option random:4 is
455 given, fio will switch to a new random file after 4 ios
456 have been issued.
457
71bfa161
JA
458ioengine=str Defines how the job issues io to the file. The following
459 types are defined:
460
461 sync Basic read(2) or write(2) io. lseek(2) is
462 used to position the io location.
463
a31041ea 464 psync Basic pread(2) or pwrite(2) io.
465
e05af9e5 466 vsync Basic readv(2) or writev(2) IO.
1d2af02a 467
15d182aa
JA
468 libaio Linux native asynchronous io. Note that Linux
469 may only support queued behaviour with
470 non-buffered IO (set direct=1 or buffered=0).
71bfa161
JA
471
472 posixaio glibc posix asynchronous io.
473
417f0068
JA
474 solarisaio Solaris native asynchronous io.
475
71bfa161
JA
476 mmap File is memory mapped and data copied
477 to/from using memcpy(3).
478
479 splice splice(2) is used to transfer the data and
480 vmsplice(2) to transfer data from user
481 space to the kernel.
482
d0ff85df
JA
483 syslet-rw Use the syslet system calls to make
484 regular read/write async.
485
71bfa161 486 sg SCSI generic sg v3 io. May either be
6c219763 487 synchronous using the SG_IO ioctl, or if
71bfa161
JA
488 the target is an sg character device
489 we use read(2) and write(2) for asynchronous
490 io.
491
a94ea28b
JA
492 null Doesn't transfer any data, just pretends
493 to. This is mainly used to exercise fio
494 itself and for debugging/testing purposes.
495
ed92ac0c
JA
496 net Transfer over the network to given host:port.
497 'filename' must be set appropriately to
414c2a3e 498 filename=host/port/protocol regardless of send
ed92ac0c 499 or receive, if the latter only the port
414c2a3e
JA
500 argument is used. 'host' may be an IP address
501 or hostname, port is the port number to be used,
502 and protocol may be 'udp' or 'tcp'. If no
503 protocol is given, TCP is used.
ed92ac0c 504
9cce02e8
JA
505 netsplice Like net, but uses splice/vmsplice to
506 map data and send/receive.
507
53aec0a4 508 cpuio Doesn't transfer any data, but burns CPU
ba0fbe10
JA
509 cycles according to the cpuload= and
510 cpucycle= options. Setting cpuload=85
511 will cause that job to do nothing but burn
36ecec83
GP
512 85% of the CPU. In case of SMP machines,
513 use numjobs=<no_of_cpu> to get desired CPU
514 usage, as the cpuload only loads a single
515 CPU at the desired rate.
ba0fbe10 516
e9a1806f
JA
517 guasi The GUASI IO engine is the Generic Userspace
518 Asyncronous Syscall Interface approach
519 to async IO. See
520
521 http://www.xmailserver.org/guasi-lib.html
522
523 for more info on GUASI.
524
8a7bd877
JA
525 external Prefix to specify loading an external
526 IO engine object file. Append the engine
527 filename, eg ioengine=external:/tmp/foo.o
528 to load ioengine foo.o in /tmp.
529
71bfa161
JA
530iodepth=int This defines how many io units to keep in flight against
531 the file. The default is 1 for each file defined in this
532 job, can be overridden with a larger value for higher
533 concurrency.
534
4950421a 535iodepth_batch_submit=int
cb5ab512 536iodepth_batch=int This defines how many pieces of IO to submit at once.
89e820f6
JA
537 It defaults to 1 which means that we submit each IO
538 as soon as it is available, but can be raised to submit
539 bigger batches of IO at the time.
cb5ab512 540
4950421a
JA
541iodepth_batch_complete=int This defines how many pieces of IO to retrieve
542 at once. It defaults to 1 which means that we'll ask
543 for a minimum of 1 IO in the retrieval process from
544 the kernel. The IO retrieval will go on until we
545 hit the limit set by iodepth_low. If this variable is
546 set to 0, then fio will always check for completed
547 events before queuing more IO. This helps reduce
548 IO latency, at the cost of more retrieval system calls.
549
e916b390
JA
550iodepth_low=int The low water mark indicating when to start filling
551 the queue again. Defaults to the same as iodepth, meaning
552 that fio will attempt to keep the queue full at all times.
553 If iodepth is set to eg 16 and iodepth_low is set to 4, then
554 after fio has filled the queue of 16 requests, it will let
555 the depth drain down to 4 before starting to fill it again.
556
71bfa161 557direct=bool If value is true, use non-buffered io. This is usually
76a43db4
JA
558 O_DIRECT.
559
560buffered=bool If value is true, use buffered io. This is the opposite
561 of the 'direct' option. Defaults to true.
71bfa161 562
f7fa2653 563offset=int Start io at the given offset in the file. The data before
71bfa161
JA
564 the given offset will not be touched. This effectively
565 caps the file size at real_size - offset.
566
567fsync=int If writing to a file, issue a sync of the dirty data
568 for every number of blocks given. For example, if you give
569 32 as a parameter, fio will sync the file for every 32
570 writes issued. If fio is using non-buffered io, we may
571 not sync the file. The exception is the sg io engine, which
6c219763 572 synchronizes the disk cache anyway.
71bfa161 573
e76b1da4 574fdatasync=int Like fsync= but uses fdatasync() to only sync data and not
5f9099ea 575 metadata blocks.
e72fa4d4
JA
576 In FreeBSD there is no fdatasync(), this falls back to
577 using fsync()
5f9099ea 578
e76b1da4
JA
579sync_file_range=str:val Use sync_file_range() for every 'val' number of
580 write operations. Fio will track range of writes that
581 have happened since the last sync_file_range() call. 'str'
582 can currently be one or more of:
583
584 wait_before SYNC_FILE_RANGE_WAIT_BEFORE
585 write SYNC_FILE_RANGE_WRITE
586 wait_after SYNC_FILE_RANGE_WAIT_AFTER
587
588 So if you do sync_file_range=wait_before,write:8, fio would
589 use SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE for
590 every 8 writes. Also see the sync_file_range(2) man page.
591 This option is Linux specific.
592
5036fc1e
JA
593overwrite=bool If true, writes to a file will always overwrite existing
594 data. If the file doesn't already exist, it will be
595 created before the write phase begins. If the file exists
596 and is large enough for the specified write phase, nothing
597 will be done.
71bfa161
JA
598
599end_fsync=bool If true, fsync file contents when the job exits.
600
ebb1415f
JA
601fsync_on_close=bool If true, fio will fsync() a dirty file on close.
602 This differs from end_fsync in that it will happen on every
603 file close, not just at the end of the job.
604
71bfa161
JA
605rwmixread=int How large a percentage of the mix should be reads.
606
607rwmixwrite=int How large a percentage of the mix should be writes. If both
608 rwmixread and rwmixwrite is given and the values do not add
609 up to 100%, the latter of the two will be used to override
c35dd7a6
JA
610 the first. This may interfere with a given rate setting,
611 if fio is asked to limit reads or writes to a certain rate.
612 If that is the case, then the distribution may be skewed.
71bfa161 613
bb8895e0
JA
614norandommap Normally fio will cover every block of the file when doing
615 random IO. If this option is given, fio will just get a
616 new random offset without looking at past io history. This
617 means that some blocks may not be read or written, and that
618 some blocks may be read/written more than once. This option
8347239a
JA
619 is mutually exclusive with verify= if and only if multiple
620 blocksizes (via bsrange=) are used, since fio only tracks
621 complete rewrites of blocks.
bb8895e0 622
2b386d25
JA
623softrandommap See norandommap. If fio runs with the random block map enabled
624 and it fails to allocate the map, if this option is set it
625 will continue without a random block map. As coverage will
626 not be as complete as with random maps, this option is
627 disabled by default.
628
71bfa161
JA
629nice=int Run the job with the given nice value. See man nice(2).
630
631prio=int Set the io priority value of this job. Linux limits us to
632 a positive value between 0 and 7, with 0 being the highest.
633 See man ionice(1).
634
635prioclass=int Set the io priority class. See man ionice(1).
636
637thinktime=int Stall the job x microseconds after an io has completed before
638 issuing the next. May be used to simulate processing being
48097d5c
JA
639 done by an application. See thinktime_blocks and
640 thinktime_spin.
641
642thinktime_spin=int
643 Only valid if thinktime is set - pretend to spend CPU time
644 doing something with the data received, before falling back
645 to sleeping for the rest of the period specified by
646 thinktime.
9c1f7434
JA
647
648thinktime_blocks
649 Only valid if thinktime is set - control how many blocks
650 to issue, before waiting 'thinktime' usecs. If not set,
651 defaults to 1 which will make fio wait 'thinktime' usecs
652 after every block.
71bfa161 653
581e7141 654rate=int Cap the bandwidth used by this job. The number is in bytes/sec,
b09da8fa 655 the normal suffix rules apply. You can use rate=500k to limit
581e7141
JA
656 reads and writes to 500k each, or you can specify read and
657 writes separately. Using rate=1m,500k would limit reads to
658 1MB/sec and writes to 500KB/sec. Capping only reads or
659 writes can be done with rate=,500k or rate=500k,. The former
660 will only limit writes (to 500KB/sec), the latter will only
661 limit reads.
71bfa161
JA
662
663ratemin=int Tell fio to do whatever it can to maintain at least this
4e991c23 664 bandwidth. Failing to meet this requirement, will cause
581e7141
JA
665 the job to exit. The same format as rate is used for
666 read vs write separation.
4e991c23
JA
667
668rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
669 as rate, just specified independently of bandwidth. If the
670 job is given a block size range instead of a fixed value,
581e7141
JA
671 the smallest block size is used as the metric. The same format
672 as rate is used for read vs write seperation.
4e991c23
JA
673
674rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
581e7141
JA
675 the job to exit. The same format as rate is used for read vs
676 write seperation.
71bfa161
JA
677
678ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
6c219763 679 of milliseconds.
71bfa161
JA
680
681cpumask=int Set the CPU affinity of this job. The parameter given is a
a08bc17f
JA
682 bitmask of allowed CPU's the job may run on. So if you want
683 the allowed CPUs to be 1 and 5, you would pass the decimal
684 value of (1 << 1 | 1 << 5), or 34. See man
7dbb6eba 685 sched_setaffinity(2). This may not work on all supported
b0ea08ce
JA
686 operating systems or kernel versions. This option doesn't
687 work well for a higher CPU count than what you can store in
688 an integer mask, so it can only control cpus 1-32. For
689 boxes with larger CPU counts, use cpus_allowed.
71bfa161 690
d2e268b0
JA
691cpus_allowed=str Controls the same options as cpumask, but it allows a text
692 setting of the permitted CPUs instead. So to use CPUs 1 and
62a7273d
JA
693 5, you would specify cpus_allowed=1,5. This options also
694 allows a range of CPUs. Say you wanted a binding to CPUs
695 1, 5, and 8-15, you would set cpus_allowed=1,5,8-15.
d2e268b0 696
e417fd66 697startdelay=time Start this job the specified number of seconds after fio
71bfa161
JA
698 has started. Only useful if the job file contains several
699 jobs, and you want to delay starting some jobs to a certain
700 time.
701
e417fd66 702runtime=time Tell fio to terminate processing after the specified number
71bfa161
JA
703 of seconds. It can be quite hard to determine for how long
704 a specified job will run, so this parameter is handy to
705 cap the total runtime to a given time.
706
cf4464ca 707time_based If set, fio will run for the duration of the runtime
bf9a3edb 708 specified even if the file(s) are completely read or
cf4464ca
JA
709 written. It will simply loop over the same workload
710 as many times as the runtime allows.
711
e417fd66 712ramp_time=time If set, fio will run the specified workload for this amount
721938ae
JA
713 of time before logging any performance numbers. Useful for
714 letting performance settle before logging results, thus
b29ee5b3
JA
715 minimizing the runtime required for stable results. Note
716 that the ramp_time is considered lead in time for a job,
717 thus it will increase the total runtime if a special timeout
718 or runtime is specified.
721938ae 719
71bfa161
JA
720invalidate=bool Invalidate the buffer/page cache parts for this file prior
721 to starting io. Defaults to true.
722
723sync=bool Use sync io for buffered writes. For the majority of the
724 io engines, this means using O_SYNC.
725
d3aad8f2 726iomem=str
71bfa161
JA
727mem=str Fio can use various types of memory as the io unit buffer.
728 The allowed values are:
729
730 malloc Use memory from malloc(3) as the buffers.
731
732 shm Use shared memory as the buffers. Allocated
733 through shmget(2).
734
74b025b0
JA
735 shmhuge Same as shm, but use huge pages as backing.
736
313cb206
JA
737 mmap Use mmap to allocate buffers. May either be
738 anonymous memory, or can be file backed if
739 a filename is given after the option. The
740 format is mem=mmap:/path/to/file.
71bfa161 741
d0bdaf49
JA
742 mmaphuge Use a memory mapped huge file as the buffer
743 backing. Append filename after mmaphuge, ala
744 mem=mmaphuge:/hugetlbfs/file
745
71bfa161 746 The area allocated is a function of the maximum allowed
5394ae5f
JA
747 bs size for the job, multiplied by the io depth given. Note
748 that for shmhuge and mmaphuge to work, the system must have
749 free huge pages allocated. This can normally be checked
750 and set by reading/writing /proc/sys/vm/nr_hugepages on a
b22989b9 751 Linux system. Fio assumes a huge page is 4MB in size. So
5394ae5f
JA
752 to calculate the number of huge pages you need for a given
753 job file, add up the io depth of all jobs (normally one unless
754 iodepth= is used) and multiply by the maximum bs set. Then
755 divide that number by the huge page size. You can see the
756 size of the huge pages in /proc/meminfo. If no huge pages
757 are allocated by having a non-zero number in nr_hugepages,
56bb17f2 758 using mmaphuge or shmhuge will fail. Also see hugepage-size.
5394ae5f
JA
759
760 mmaphuge also needs to have hugetlbfs mounted and the file
761 location should point there. So if it's mounted in /huge,
762 you would use mem=mmaphuge:/huge/somefile.
71bfa161 763
d529ee19
JA
764iomem_align=int This indiciates the memory alignment of the IO memory buffers.
765 Note that the given alignment is applied to the first IO unit
766 buffer, if using iodepth the alignment of the following buffers
767 are given by the bs used. In other words, if using a bs that is
768 a multiple of the page sized in the system, all buffers will
769 be aligned to this value. If using a bs that is not page
770 aligned, the alignment of subsequent IO memory buffers is the
771 sum of the iomem_align and bs used.
772
f7fa2653 773hugepage-size=int
56bb17f2 774 Defines the size of a huge page. Must at least be equal
b22989b9 775 to the system setting, see /proc/meminfo. Defaults to 4MB.
c51074e7
JA
776 Should probably always be a multiple of megabytes, so using
777 hugepage-size=Xm is the preferred way to set this to avoid
778 setting a non-pow-2 bad value.
56bb17f2 779
71bfa161
JA
780exitall When one job finishes, terminate the rest. The default is
781 to wait for each job to finish, sometimes that is not the
782 desired action.
783
784bwavgtime=int Average the calculated bandwidth over the given time. Value
6c219763 785 is specified in milliseconds.
71bfa161
JA
786
787create_serialize=bool If true, serialize the file creating for the jobs.
788 This may be handy to avoid interleaving of data
789 files, which may greatly depend on the filesystem
790 used and even the number of processors in the system.
791
792create_fsync=bool fsync the data file after creation. This is the
793 default.
794
814452bd
JA
795create_on_open=bool Don't pre-setup the files for IO, just create open()
796 when it's time to do IO to that file.
797
afad68f7 798pre_read=bool If this is given, files will be pre-read into memory before
34f1c044
JA
799 starting the given IO operation. This will also clear
800 the 'invalidate' flag, since it is pointless to pre-read
9c0d2241
JA
801 and then drop the cache. This will only work for IO engines
802 that are seekable, since they allow you to read the same data
803 multiple times. Thus it will not work on eg network or splice
804 IO.
afad68f7 805
e545a6ce 806unlink=bool Unlink the job files when done. Not the default, as repeated
bf9a3edb
JA
807 runs of that job would then waste time recreating the file
808 set again and again.
71bfa161
JA
809
810loops=int Run the specified number of iterations of this job. Used
811 to repeat the same workload a given number of times. Defaults
812 to 1.
813
68e1f29a 814do_verify=bool Run the verify phase after a write phase. Only makes sense if
e84c73a8
SL
815 verify is set. Defaults to 1.
816
71bfa161
JA
817verify=str If writing to a file, fio can verify the file contents
818 after each iteration of the job. The allowed values are:
819
820 md5 Use an md5 sum of the data area and store
821 it in the header of each block.
822
17dc34df
JA
823 crc64 Use an experimental crc64 sum of the data
824 area and store it in the header of each
825 block.
826
bac39e0e
JA
827 crc32c Use a crc32c sum of the data area and store
828 it in the header of each block.
829
3845591f 830 crc32c-intel Use hardware assisted crc32c calcuation
0539d758
JA
831 provided on SSE4.2 enabled processors. Falls
832 back to regular software crc32c, if not
833 supported by the system.
3845591f 834
71bfa161
JA
835 crc32 Use a crc32 sum of the data area and store
836 it in the header of each block.
837
969f7ed3
JA
838 crc16 Use a crc16 sum of the data area and store
839 it in the header of each block.
840
17dc34df
JA
841 crc7 Use a crc7 sum of the data area and store
842 it in the header of each block.
843
cd14cc10
JA
844 sha512 Use sha512 as the checksum function.
845
846 sha256 Use sha256 as the checksum function.
847
7c353ceb
JA
848 sha1 Use optimized sha1 as the checksum function.
849
7437ee87
SL
850 meta Write extra information about each io
851 (timestamp, block number etc.). The block
996093bb 852 number is verified. See also verify_pattern.
7437ee87 853
36690c9b
JA
854 null Only pretend to verify. Useful for testing
855 internals with ioengine=null, not for much
856 else.
857
6c219763 858 This option can be used for repeated burn-in tests of a
71bfa161 859 system to make sure that the written data is also
b892dc08
JA
860 correctly read back. If the data direction given is
861 a read or random read, fio will assume that it should
862 verify a previously written file. If the data direction
863 includes any form of write, the verify will be of the
864 newly written data.
71bfa161 865
160b966d
JA
866verifysort=bool If set, fio will sort written verify blocks when it deems
867 it faster to read them back in a sorted manner. This is
868 often the case when overwriting an existing file, since
869 the blocks are already laid out in the file system. You
870 can ignore this option unless doing huge amounts of really
871 fast IO where the red-black tree sorting CPU time becomes
872 significant.
3f9f4e26 873
f7fa2653 874verify_offset=int Swap the verification header with data somewhere else
546a9142
SL
875 in the block before writing. Its swapped back before
876 verifying.
877
f7fa2653 878verify_interval=int Write the verification header at a finer granularity
3f9f4e26
SL
879 than the blocksize. It will be written for chunks the
880 size of header_interval. blocksize should divide this
881 evenly.
90059d65 882
0e92f873 883verify_pattern=str If set, fio will fill the io buffers with this
e28218f3
SL
884 pattern. Fio defaults to filling with totally random
885 bytes, but sometimes it's interesting to fill with a known
886 pattern for io verification purposes. Depending on the
887 width of the pattern, fio will fill 1/2/3/4 bytes of the
0e92f873
RR
888 buffer at the time(it can be either a decimal or a hex number).
889 The verify_pattern if larger than a 32-bit quantity has to
996093bb
JA
890 be a hex number that starts with either "0x" or "0X". Use
891 with verify=meta.
e28218f3 892
68e1f29a 893verify_fatal=bool Normally fio will keep checking the entire contents
a12a3b4d
JA
894 before quitting on a block verification failure. If this
895 option is set, fio will exit the job on the first observed
896 failure.
e8462bd8
JA
897
898verify_async=int Fio will normally verify IO inline from the submitting
899 thread. This option takes an integer describing how many
900 async offload threads to create for IO verification instead,
901 causing fio to offload the duty of verifying IO contents
c85c324c
JA
902 to one or more separate threads. If using this offload
903 option, even sync IO engines can benefit from using an
904 iodepth setting higher than 1, as it allows them to have
905 IO in flight while verifies are running.
e8462bd8
JA
906
907verify_async_cpus=str Tell fio to set the given CPU affinity on the
908 async IO verification threads. See cpus_allowed for the
909 format used.
6f87418f
JA
910
911verify_backlog=int Fio will normally verify the written contents of a
912 job that utilizes verify once that job has completed. In
913 other words, everything is written then everything is read
914 back and verified. You may want to verify continually
915 instead for a variety of reasons. Fio stores the meta data
916 associated with an IO block in memory, so for large
917 verify workloads, quite a bit of memory would be used up
918 holding this meta data. If this option is enabled, fio
919 will verify the previously written blocks before continuing
920 to write new ones.
921
922verify_backlog_batch=int Control how many blocks fio will verify
923 if verify_backlog is set. If not set, will default to
924 the value of verify_backlog (meaning the entire queue
925 is read back and verified).
160b966d 926
71bfa161
JA
927stonewall Wait for preceeding jobs in the job file to exit, before
928 starting this one. Can be used to insert serialization
b3d62a75
JA
929 points in the job file. A stone wall also implies starting
930 a new reporting group.
931
932new_group Start a new reporting group. If this option isn't given,
933 jobs in a file will be part of the same reporting group
bf9a3edb 934 unless separated by a stone wall (or if it's a group
b3d62a75 935 by itself, with the numjobs option).
71bfa161
JA
936
937numjobs=int Create the specified number of clones of this job. May be
938 used to setup a larger number of threads/processes doing
fa28c85a
JA
939 the same thing. We regard that grouping of jobs as a
940 specific group.
941
942group_reporting If 'numjobs' is set, it may be interesting to display
943 statistics for the group as a whole instead of for each
944 individual job. This is especially true of 'numjobs' is
945 large, looking at individual thread/process output quickly
946 becomes unwieldy. If 'group_reporting' is specified, fio
947 will show the final report per-group instead of per-job.
71bfa161
JA
948
949thread fio defaults to forking jobs, however if this option is
950 given, fio will use pthread_create(3) to create threads
951 instead.
952
f7fa2653 953zonesize=int Divide a file into zones of the specified size. See zoneskip.
71bfa161 954
f7fa2653 955zoneskip=int Skip the specified number of bytes when zonesize data has
71bfa161
JA
956 been read. The two zone options can be used to only do
957 io on zones of a file.
958
076efc7c
JA
959write_iolog=str Write the issued io patterns to the specified file. See
960 read_iolog.
71bfa161 961
076efc7c 962read_iolog=str Open an iolog with the specified file name and replay the
71bfa161 963 io patterns it contains. This can be used to store a
6df8adaa
JA
964 workload and replay it sometime later. The iolog given
965 may also be a blktrace binary file, which allows fio
966 to replay a workload captured by blktrace. See blktrace
967 for how to capture such logging data. For blktrace replay,
968 the file needs to be turned into a blkparse binary data
ea3e51c3 969 file first (blkparse <device> -o /dev/null -d file_for_fio.bin).
71bfa161 970
e3cedca7 971write_bw_log=str If given, write a bandwidth log of the jobs in this job
71bfa161 972 file. Can be used to store data of the bandwidth of the
e0da9bc2
JA
973 jobs in their lifetime. The included fio_generate_plots
974 script uses gnuplot to turn these text files into nice
e3cedca7
JA
975 graphs. See write_log_log for behaviour of given
976 filename. For this option, the postfix is _bw.log.
71bfa161 977
e3cedca7
JA
978write_lat_log=str Same as write_bw_log, except that this option stores io
979 completion latencies instead. If no filename is given
980 with this option, the default filename of "jobname_type.log"
981 is used. Even if the filename is given, fio will still
982 append the type of log. So if one specifies
983
984 write_lat_log=foo
985
986 The actual log names will be foo_clat.log and foo_slat.log.
987 This helps fio_generate_plot fine the logs automatically.
71bfa161 988
f7fa2653 989lockmem=int Pin down the specified amount of memory with mlock(2). Can
71bfa161
JA
990 potentially be used instead of removing memory or booting
991 with less memory to simulate a smaller amount of memory.
992
993exec_prerun=str Before running this job, issue the command specified
994 through system(3).
995
996exec_postrun=str After the job completes, issue the command specified
997 though system(3).
998
999ioscheduler=str Attempt to switch the device hosting the file to the specified
1000 io scheduler before running.
1001
1002cpuload=int If the job is a CPU cycle eater, attempt to use the specified
1003 percentage of CPU cycles.
1004
1005cpuchunks=int If the job is a CPU cycle eater, split the load into
26eca2db 1006 cycles of the given time. In microseconds.
71bfa161 1007
0a839f30
JA
1008disk_util=bool Generate disk utilization statistics, if the platform
1009 supports it. Defaults to on.
1010
9520ebb9
JA
1011disable_clat=bool Disable measurements of completion latency numbers. Useful
1012 only for cutting back the number of calls to gettimeofday,
1013 as that does impact performance at really high IOPS rates.
1014 Note that to really get rid of a large amount of these
1015 calls, this option must be used with disable_slat and
1016 disable_bw as well.
1017
1018disable_slat=bool Disable measurements of submission latency numbers. See
1019 disable_clat.
1020
1021disable_bw=bool Disable measurements of throughput/bandwidth numbers. See
1022 disable_clat.
1023
993bf48b
JA
1024gtod_reduce=bool Enable all of the gettimeofday() reducing options
1025 (disable_clat, disable_slat, disable_bw) plus reduce
1026 precision of the timeout somewhat to really shrink
1027 the gettimeofday() call count. With this option enabled,
1028 we only do about 0.4% of the gtod() calls we would have
1029 done if all time keeping was enabled.
1030
be4ecfdf
JA
1031gtod_cpu=int Sometimes it's cheaper to dedicate a single thread of
1032 execution to just getting the current time. Fio (and
1033 databases, for instance) are very intensive on gettimeofday()
1034 calls. With this option, you can set one CPU aside for
1035 doing nothing but logging current time to a shared memory
1036 location. Then the other threads/processes that run IO
1037 workloads need only copy that segment, instead of entering
1038 the kernel with a gettimeofday() call. The CPU set aside
1039 for doing these time calls will be excluded from other
1040 uses. Fio will manually clear it from the CPU mask of other
1041 jobs.
a696fa2a 1042
f2bba182
RR
1043continue_on_error=bool Normally fio will exit the job on the first observed
1044 failure. If this option is set, fio will continue the job when
1045 there is a 'non-fatal error' (EIO or EILSEQ) until the runtime
1046 is exceeded or the I/O size specified is completed. If this
1047 option is used, there are two more stats that are appended,
1048 the total error count and the first error. The error field
1049 given in the stats is the first error that was hit during the
1050 run.
be4ecfdf 1051
6adb38a1
JA
1052cgroup=str Add job to this control group. If it doesn't exist, it will
1053 be created. The system must have a mounted cgroup blkio
1054 mount point for this to work. If your system doesn't have it
1055 mounted, you can do so with:
a696fa2a
JA
1056
1057 # mount -t cgroup -o blkio none /cgroup
1058
a696fa2a
JA
1059cgroup_weight=int Set the weight of the cgroup to this value. See
1060 the documentation that comes with the kernel, allowed values
1061 are in the range of 100..1000.
71bfa161 1062
7de87099
VG
1063cgroup_nodelete=bool Normally fio will delete the cgroups it has created after
1064 the job completion. To override this behavior and to leave
1065 cgroups around after the job completion, set cgroup_nodelete=1.
1066 This can be useful if one wants to inspect various cgroup
1067 files after job completion. Default: false
1068
e0b0d892
JA
1069uid=int Instead of running as the invoking user, set the user ID to
1070 this value before the thread/process does any work.
1071
1072gid=int Set group ID, see uid.
1073
71bfa161
JA
10746.0 Interpreting the output
1075---------------------------
1076
1077fio spits out a lot of output. While running, fio will display the
1078status of the jobs created. An example of that would be:
1079
73c8b082 1080Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
71bfa161
JA
1081
1082The characters inside the square brackets denote the current status of
1083each thread. The possible values (in typical life cycle order) are:
1084
1085Idle Run
1086---- ---
1087P Thread setup, but not started.
1088C Thread created.
1089I Thread initialized, waiting.
b0f65863 1090 p Thread running pre-reading file(s).
71bfa161
JA
1091 R Running, doing sequential reads.
1092 r Running, doing random reads.
1093 W Running, doing sequential writes.
1094 w Running, doing random writes.
1095 M Running, doing mixed sequential reads/writes.
1096 m Running, doing mixed random reads/writes.
1097 F Running, currently waiting for fsync()
fc6bd43c 1098 V Running, doing verification of written data.
71bfa161
JA
1099E Thread exited, not reaped by main thread yet.
1100_ Thread reaped.
1101
1102The other values are fairly self explanatory - number of threads
c9f60304
JA
1103currently running and doing io, rate of io since last check (read speed
1104listed first, then write speed), and the estimated completion percentage
1105and time for the running group. It's impossible to estimate runtime of
1106the following groups (if any).
71bfa161
JA
1107
1108When fio is done (or interrupted by ctrl-c), it will show the data for
1109each thread, group of threads, and disks in that order. For each data
1110direction, the output looks like:
1111
1112Client1 (g=0): err= 0:
b22989b9 1113 write: io= 32MB, bw= 666KB/s, runt= 50320msec
6104ddb6
JA
1114 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
1115 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
b22989b9 1116 bw (KB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
e7823a94 1117 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
71619dc2 1118 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
1119 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1120 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
30061b97 1121 issued r/w: total=0/32768, short=0/0
8abdce66
JA
1122 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
1123 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
71bfa161
JA
1124
1125The client number is printed, along with the group id and error of that
1126thread. Below is the io statistics, here for writes. In the order listed,
1127they denote:
1128
1129io= Number of megabytes io performed
1130bw= Average bandwidth rate
1131runt= The runtime of that thread
72fbda2a 1132 slat= Submission latency (avg being the average, stdev being the
71bfa161
JA
1133 standard deviation). This is the time it took to submit
1134 the io. For sync io, the slat is really the completion
8a35c71e 1135 latency, since queue/complete is one operation there. This
bf9a3edb 1136 value can be in milliseconds or microseconds, fio will choose
8a35c71e 1137 the most appropriate base and print that. In the example
bf9a3edb 1138 above, milliseconds is the best scale.
71bfa161
JA
1139 clat= Completion latency. Same names as slat, this denotes the
1140 time from submission to completion of the io pieces. For
1141 sync io, clat will usually be equal (or very close) to 0,
1142 as the time from submit to complete is basically just
1143 CPU time (io has already been done, see slat explanation).
1144 bw= Bandwidth. Same names as the xlat stats, but also includes
1145 an approximate percentage of total aggregate bandwidth
1146 this thread received in this group. This last value is
1147 only really useful if the threads in this group are on the
1148 same disk, since they are then competing for disk access.
1149cpu= CPU usage. User and system time, along with the number
e7823a94
JA
1150 of context switches this thread went through, usage of
1151 system and user time, and finally the number of major
1152 and minor page faults.
71619dc2
JA
1153IO depths= The distribution of io depths over the job life time. The
1154 numbers are divided into powers of 2, so for example the
1155 16= entries includes depths up to that value but higher
1156 than the previous entry. In other words, it covers the
1157 range from 16 to 31.
838bc709
JA
1158IO submit= How many pieces of IO were submitting in a single submit
1159 call. Each entry denotes that amount and below, until
1160 the previous entry - eg, 8=100% mean that we submitted
1161 anywhere in between 5-8 ios per submit call.
1162IO complete= Like the above submit number, but for completions instead.
30061b97
JA
1163IO issued= The number of read/write requests issued, and how many
1164 of them were short.
ec118304
JA
1165IO latencies= The distribution of IO completion latencies. This is the
1166 time from when IO leaves fio and when it gets completed.
1167 The numbers follow the same pattern as the IO depths,
1168 meaning that 2=1.6% means that 1.6% of the IO completed
8abdce66
JA
1169 within 2 msecs, 20=12.8% means that 12.8% of the IO
1170 took more than 10 msecs, but less than (or equal to) 20 msecs.
71bfa161
JA
1171
1172After each client has been listed, the group statistics are printed. They
1173will look like this:
1174
1175Run status group 0 (all jobs):
b22989b9
JA
1176 READ: io=64MB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
1177 WRITE: io=64MB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
71bfa161
JA
1178
1179For each data direction, it prints:
1180
1181io= Number of megabytes io performed.
1182aggrb= Aggregate bandwidth of threads in this group.
1183minb= The minimum average bandwidth a thread saw.
1184maxb= The maximum average bandwidth a thread saw.
1185mint= The smallest runtime of the threads in that group.
1186maxt= The longest runtime of the threads in that group.
1187
1188And finally, the disk statistics are printed. They will look like this:
1189
1190Disk stats (read/write):
1191 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
1192
1193Each value is printed for both reads and writes, with reads first. The
1194numbers denote:
1195
1196ios= Number of ios performed by all groups.
1197merge= Number of merges io the io scheduler.
1198ticks= Number of ticks we kept the disk busy.
1199io_queue= Total time spent in the disk queue.
1200util= The disk utilization. A value of 100% means we kept the disk
1201 busy constantly, 50% would be a disk idling half of the time.
1202
1203
12047.0 Terse output
1205----------------
1206
1207For scripted usage where you typically want to generate tables or graphs
6af019c9 1208of the results, fio can output the results in a semicolon separated format.
71bfa161
JA
1209The format is one long line of values, such as:
1210
6af019c9
JA
1211client1;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%
1212;0.0%;0.0%;0.0%;0.0%;0.0%
71bfa161 1213
6820cb3b
JA
1214To enable terse output, use the --minimal command line option.
1215
71bfa161
JA
1216Split up, the format is as follows:
1217
1218 jobname, groupid, error
1219 READ status:
b22989b9 1220 KB IO, bandwidth (KB/sec), runtime (msec)
71bfa161
JA
1221 Submission latency: min, max, mean, deviation
1222 Completion latency: min, max, mean, deviation
6c219763 1223 Bw: min, max, aggregate percentage of total, mean, deviation
71bfa161 1224 WRITE status:
b22989b9 1225 KB IO, bandwidth (KB/sec), runtime (msec)
71bfa161
JA
1226 Submission latency: min, max, mean, deviation
1227 Completion latency: min, max, mean, deviation
6c219763 1228 Bw: min, max, aggregate percentage of total, mean, deviation
046ee302 1229 CPU usage: user, system, context switches, major faults, minor faults
2270890c
JA
1230 IO depths: <=1, 2, 4, 8, 16, 32, >=64
1231 IO latencies: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, >=2000
1232 Text description
71bfa161 1233