8 5. Detailed list of parameters
13 1.0 Overview and history
14 ------------------------
15 fio was originally written to save me the hassle of writing special test
16 case programs when I wanted to test a specific workload, either for
17 performance reasons or to find/reproduce a bug. The process of writing
18 such a test app can be tiresome, especially if you have to do it often.
19 Hence I needed a tool that would be able to simulate a given io workload
20 without resorting to writing a tailored test case again and again.
22 A test work load is difficult to define, though. There can be any number
23 of processes or threads involved, and they can each be using their own
24 way of generating io. You could have someone dirtying large amounts of
25 memory in an memory mapped file, or maybe several threads issuing
26 reads using asynchronous io. fio needed to be flexible enough to
27 simulate both of these cases, and many more.
31 The first step in getting fio to simulate a desired io workload, is
32 writing a job file describing that specific setup. A job file may contain
33 any number of threads and/or files - the typical contents of the job file
34 is a global section defining shared parameters, and one or more job
35 sections describing the jobs involved. When run, fio parses this file
36 and sets everything up as described. If we break down a job from top to
37 bottom, it contains the following basic parameters:
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.
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
48 IO size How much data are we going to be reading/writing.
50 IO engine How do we issue io? We could be memory mapping the
51 file, we could be using regular read/write, we
52 could be using splice, async io, syslet, or even
55 IO depth If the io engine is async, how large a queuing
56 depth do we want to maintain?
58 IO type Should we be doing buffered io, or direct/raw io?
60 Num files How many files are we spreading the workload over.
62 Num threads How many threads or processes should we spread
65 The above are the basic parameters defined for a workload, in addition
66 there's a multitude of parameters that modify other aspects of how this
72 See the README file for command line parameters, there are only a few
75 Running fio is normally the easiest part - you just give it the job file
76 (or job files) as parameters:
80 and it will start doing what the job_file tells it to do. You can give
81 more than one job file on the command line, fio will serialize the running
82 of those files. Internally that is the same as using the 'stonewall'
83 parameter described the the parameter section.
85 If the job file contains only one job, you may as well just give the
86 parameters on the command line. The command line parameters are identical
87 to the job parameters, with a few extra that control global parameters
88 (see README). For example, for the job file parameter iodepth=2, the
89 mirror command line option would be --iodepth 2 or --iodepth=2. You can
90 also 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.
92 Command line entries following a --name entry will apply to that job,
93 until there are no more entries or a new --name entry is seen. This is
94 similar to the job file options, where each option applies to the current
95 job until a new [] job entry is seen.
97 fio does not need to run as root, except if the files or devices specified
98 in the job section requires that. Some other options may also be restricted,
99 such as memory locking, io scheduler switching, and decreasing the nice value.
104 As previously described, fio accepts one or more job files describing
105 what it is supposed to do. The job file format is the classic ini file,
106 where the names enclosed in [] brackets define the job name. You are free
107 to use any ascii name you want, except 'global' which has special meaning.
108 A global section sets defaults for the jobs described in that file. A job
109 may override a global section parameter, and a job file may even have
110 several global sections if so desired. A job is only affected by a global
111 section residing above it. If the first character in a line is a ';' or a
112 '#', the entire line is discarded as a comment.
114 So let's look at a really simple job file that defines two processes, each
115 randomly reading from a 128MB file.
117 ; -- start job file --
128 As you can see, the job file sections themselves are empty as all the
129 described parameters are shared. As no filename= option is given, fio
130 makes up a filename for each of the jobs as it sees fit. On the command
131 line, this job would look as follows:
133 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
136 Let's look at an example that has a number of processes writing randomly
139 ; -- start job file --
151 Here we have no global section, as we only have one job defined anyway.
152 We want to use async io here, with a depth of 4 for each file. We also
153 increased the buffer size used to 32KB and define numjobs to 4 to
154 fork 4 identical jobs. The result is 4 processes each randomly writing
155 to their own 64MB file. Instead of using the above job file, you could
156 have given the parameters on the command line. For this case, you would
159 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
161 4.1 Environment variables
162 -------------------------
164 fio also supports environment variable expansion in job files. Any
165 substring of the form "${VARNAME}" as part of an option value (in other
166 words, on the right of the `='), will be expanded to the value of the
167 environment variable called VARNAME. If no such environment variable
168 is defined, or VARNAME is the empty string, the empty string will be
171 As an example, let's look at a sample fio invocation and job file:
173 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
175 ; -- start job file --
182 This will expand to the following equivalent job file at runtime:
184 ; -- start job file --
191 fio ships with a few example job files, you can also look there for
194 4.2 Reserved keywords
195 ---------------------
197 Additionally, fio has a set of reserved keywords that will be replaced
198 internally with the appropriate value. Those keywords are:
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
204 These can be used on the command line or in the job file, and will be
205 automatically substituted with the current system values when the job
206 is run. Simple math is also supported on these keywords, so you can
207 perform actions like:
211 and get that properly expanded to 8 times the size of memory in the
215 5.0 Detailed list of parameters
216 -------------------------------
218 This section describes in details each parameter associated with a job.
219 Some parameters take an option of a given type, such as an integer or
220 a string. The following types are used:
222 str String. This is a sequence of alpha characters.
223 time Integer with possible time suffix. In seconds unless otherwise
224 specified, use eg 10m for 10 minutes. Accepts s/m/h for seconds,
226 int 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
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
231 out '4096' or just give 4k. The suffixes signify base 2 values, so
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.
240 bool Boolean. Usually parsed as an integer, however only defined for
241 true and false (1 and 0).
242 irange Integer range with suffix. Allows value range to be given, such
243 as 1024-4096. A colon may also be used as the separator, eg
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
248 With the above in mind, here follows the complete list of fio job
251 name=str ASCII name of the job. This may be used to override the
252 name printed by fio for this job. Otherwise the job
253 name is used. On the command line this parameter has the
254 special purpose of also signaling the start of a new
257 description=str Text description of the job. Doesn't do anything except
258 dump this text description when this job is run. It's
261 directory=str Prefix filenames with this directory. Used to place files
262 in a different location than "./".
264 filename=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
267 a filename for each of them to override the default. If
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
274 filename=/dev/sda:/dev/sdb. On Windows, disk devices are accessed
275 as \\.\PhysicalDrive0 for the first device, \\.\PhysicalDrive1
276 for the second etc. If the wanted filename does need to
277 include a colon, then escape that with a '\' character.
278 For instance, if the filename is "/dev/dsk/foo@3,0:c",
279 then you would use filename="/dev/dsk/foo@3,0\:c".
280 '-' is a reserved name, meaning stdin or stdout. Which of the
281 two depends on the read/write direction set.
283 opendir=str Tell fio to recursively add any file it can find in this
284 directory and down the file system tree.
286 lockfile=str Fio defaults to not locking any files before it does
287 IO to them. If a file or file descriptor is shared, fio
288 can serialize IO to that file to make the end result
289 consistent. This is usual for emulating real workloads that
290 share files. The lock modes are:
292 none No locking. The default.
293 exclusive Only one thread/process may do IO,
294 excluding all others.
295 readwrite Read-write locking on the file. Many
296 readers may access the file at the
297 same time, but writes get exclusive
300 The option may be post-fixed with a lock batch number. If
301 set, then each thread/process may do that amount of IOs to
302 the file before giving up the lock. Since lock acquisition is
303 expensive, batching the lock/unlocks will speed up IO.
306 rw=str Type of io pattern. Accepted values are:
308 read Sequential reads
309 write Sequential writes
310 randwrite Random writes
311 randread Random reads
312 rw Sequential mixed reads and writes
313 randrw Random mixed reads and writes
315 For the mixed io types, the default is to split them 50/50.
316 For certain types of io the result may still be skewed a bit,
317 since the speed may be different. It is possible to specify
318 a number of IO's to do before getting a new offset, this is
319 one by appending a ':<nr>' to the end of the string given.
320 For a random read, it would look like 'rw=randread:8' for
321 passing in an offset modifier with a value of 8. See the
322 'rw_sequencer' option.
324 rw_sequencer=str If an offset modifier is given by appending a number to
325 the rw=<str> line, then this option controls how that
326 number modifies the IO offset being generated. Accepted
329 sequential Generate sequential offset
330 identical Generate the same offset
332 'sequential' is only useful for random IO, where fio would
333 normally generate a new random offset for every IO. If you
334 append eg 8 to randread, you would get a new random offset for
335 every 8 IO's. The result would be a seek for only every 8
336 IO's, instead of for every IO. Use rw=randread:8 to specify
337 that. As sequential IO is already sequential, setting
338 'sequential' for that would not result in any differences.
339 'identical' behaves in a similar fashion, except it sends
340 the same offset 8 number of times before generating a new
343 kb_base=int The base unit for a kilobyte. The defacto base is 2^10, 1024.
344 Storage manufacturers like to use 10^3 or 1000 as a base
345 ten unit instead, for obvious reasons. Allow values are
346 1024 or 1000, with 1024 being the default.
348 randrepeat=bool For random IO workloads, seed the generator in a predictable
349 way so that results are repeatable across repetitions.
351 fallocate=bool By default, fio will use fallocate() to advise the system
352 of the size of the file we are going to write. This can be
353 turned off with fallocate=0. May not be available on all
354 supported platforms. If using ZFS on Solaris this must be
355 set to 0 because ZFS doesn't support it.
357 fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
358 on what IO patterns it is likely to issue. Sometimes you
359 want to test specific IO patterns without telling the
360 kernel about it, in which case you can disable this option.
361 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
362 IO and POSIX_FADV_RANDOM for random IO.
364 size=int The total size of file io for this job. Fio will run until
365 this many bytes has been transferred, unless runtime is
366 limited by other options (such as 'runtime', for instance).
367 Unless specific nrfiles and filesize options are given,
368 fio will divide this size between the available files
369 specified by the job. If not set, fio will use the full
370 size of the given files or devices. If the the files
371 do not exist, size must be given.
373 filesize=int Individual file sizes. May be a range, in which case fio
374 will select sizes for files at random within the given range
375 and limited to 'size' in total (if that is given). If not
376 given, each created file is the same size.
379 fill_fs=bool Sets size to something really large and waits for ENOSPC (no
380 space left on device) as the terminating condition. Only makes
381 sense with sequential write. For a read workload, the mount
382 point will be filled first then IO started on the result. This
383 option doesn't make sense if operating on a raw device node,
384 since the size of that is already known by the file system.
385 Additionally, writing beyond end-of-device will not return
389 bs=int The block size used for the io units. Defaults to 4k. Values
390 can be given for both read and writes. If a single int is
391 given, it will apply to both. If a second int is specified
392 after a comma, it will apply to writes only. In other words,
393 the format is either bs=read_and_write or bs=read,write.
394 bs=4k,8k will thus use 4k blocks for reads, and 8k blocks
395 for writes. If you only wish to set the write size, you
396 can do so by passing an empty read size - bs=,8k will set
397 8k for writes and leave the read default value.
400 ba=int At what boundary to align random IO offsets. Defaults to
401 the same as 'blocksize' the minimum blocksize given.
402 Minimum alignment is typically 512b for using direct IO,
403 though it usually depends on the hardware block size. This
404 option is mutually exclusive with using a random map for
405 files, so it will turn off that option.
407 blocksize_range=irange
408 bsrange=irange Instead of giving a single block size, specify a range
409 and fio will mix the issued io block sizes. The issued
410 io unit will always be a multiple of the minimum value
411 given (also see bs_unaligned). Applies to both reads and
412 writes, however a second range can be given after a comma.
415 bssplit=str Sometimes you want even finer grained control of the
416 block sizes issued, not just an even split between them.
417 This option allows you to weight various block sizes,
418 so that you are able to define a specific amount of
419 block sizes issued. The format for this option is:
421 bssplit=blocksize/percentage:blocksize/percentage
423 for as many block sizes as needed. So if you want to define
424 a workload that has 50% 64k blocks, 10% 4k blocks, and
425 40% 32k blocks, you would write:
427 bssplit=4k/10:64k/50:32k/40
429 Ordering does not matter. If the percentage is left blank,
430 fio will fill in the remaining values evenly. So a bssplit
431 option like this one:
433 bssplit=4k/50:1k/:32k/
435 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
436 always add up to 100, if bssplit is given a range that adds
437 up to more, it will error out.
439 bssplit also supports giving separate splits to reads and
440 writes. The format is identical to what bs= accepts. You
441 have to separate the read and write parts with a comma. So
442 if you want a workload that has 50% 2k reads and 50% 4k reads,
443 while having 90% 4k writes and 10% 8k writes, you would
446 bssplit=2k/50:4k/50,4k/90,8k/10
449 bs_unaligned If this option is given, any byte size value within bsrange
450 may be used as a block range. This typically wont work with
451 direct IO, as that normally requires sector alignment.
453 zero_buffers If this option is given, fio will init the IO buffers to
454 all zeroes. The default is to fill them with random data.
456 refill_buffers If this option is given, fio will refill the IO buffers
457 on every submit. The default is to only fill it at init
458 time and reuse that data. Only makes sense if zero_buffers
459 isn't specified, naturally. If data verification is enabled,
460 refill_buffers is also automatically enabled.
462 nrfiles=int Number of files to use for this job. Defaults to 1.
464 openfiles=int Number of files to keep open at the same time. Defaults to
465 the same as nrfiles, can be set smaller to limit the number
468 file_service_type=str Defines how fio decides which file from a job to
469 service next. The following types are defined:
471 random Just choose a file at random.
473 roundrobin Round robin over open files. This
476 sequential Finish one file before moving on to
477 the next. Multiple files can still be
478 open depending on 'openfiles'.
480 The string can have a number appended, indicating how
481 often to switch to a new file. So if option random:4 is
482 given, fio will switch to a new random file after 4 ios
485 ioengine=str Defines how the job issues io to the file. The following
488 sync Basic read(2) or write(2) io. lseek(2) is
489 used to position the io location.
491 psync Basic pread(2) or pwrite(2) io.
493 vsync Basic readv(2) or writev(2) IO.
495 libaio Linux native asynchronous io. Note that Linux
496 may only support queued behaviour with
497 non-buffered IO (set direct=1 or buffered=0).
499 posixaio glibc posix asynchronous io.
501 solarisaio Solaris native asynchronous io.
503 windowsaio Windows native asynchronous io.
505 mmap File is memory mapped and data copied
506 to/from using memcpy(3).
508 splice splice(2) is used to transfer the data and
509 vmsplice(2) to transfer data from user
512 syslet-rw Use the syslet system calls to make
513 regular read/write async.
515 sg SCSI generic sg v3 io. May either be
516 synchronous using the SG_IO ioctl, or if
517 the target is an sg character device
518 we use read(2) and write(2) for asynchronous
521 null Doesn't transfer any data, just pretends
522 to. This is mainly used to exercise fio
523 itself and for debugging/testing purposes.
525 net Transfer over the network to given host:port.
526 'filename' must be set appropriately to
527 filename=host/port/protocol regardless of send
528 or receive, if the latter only the port
529 argument is used. 'host' may be an IP address
530 or hostname, port is the port number to be used,
531 and protocol may be 'udp' or 'tcp'. If no
532 protocol is given, TCP is used.
534 netsplice Like net, but uses splice/vmsplice to
535 map data and send/receive.
537 cpuio Doesn't transfer any data, but burns CPU
538 cycles according to the cpuload= and
539 cpucycle= options. Setting cpuload=85
540 will cause that job to do nothing but burn
541 85% of the CPU. In case of SMP machines,
542 use numjobs=<no_of_cpu> to get desired CPU
543 usage, as the cpuload only loads a single
544 CPU at the desired rate.
546 guasi The GUASI IO engine is the Generic Userspace
547 Asyncronous Syscall Interface approach
550 http://www.xmailserver.org/guasi-lib.html
552 for more info on GUASI.
554 external Prefix to specify loading an external
555 IO engine object file. Append the engine
556 filename, eg ioengine=external:/tmp/foo.o
557 to load ioengine foo.o in /tmp.
559 iodepth=int This defines how many io units to keep in flight against
560 the file. The default is 1 for each file defined in this
561 job, can be overridden with a larger value for higher
562 concurrency. Note that increasing iodepth beyond 1 will not
563 affect synchronous ioengines (except for small degress when
564 verify_async is in use). Even async engines may impose OS
565 restrictions causing the desired depth not to be achieved.
566 This may happen on Linux when using libaio and not setting
567 direct=1, since buffered IO is not async on that OS. Keep an
568 eye on the IO depth distribution in the fio output to verify
569 that the achieved depth is as expected. Default: 1.
571 iodepth_batch_submit=int
572 iodepth_batch=int This defines how many pieces of IO to submit at once.
573 It defaults to 1 which means that we submit each IO
574 as soon as it is available, but can be raised to submit
575 bigger batches of IO at the time.
577 iodepth_batch_complete=int This defines how many pieces of IO to retrieve
578 at once. It defaults to 1 which means that we'll ask
579 for a minimum of 1 IO in the retrieval process from
580 the kernel. The IO retrieval will go on until we
581 hit the limit set by iodepth_low. If this variable is
582 set to 0, then fio will always check for completed
583 events before queuing more IO. This helps reduce
584 IO latency, at the cost of more retrieval system calls.
586 iodepth_low=int The low water mark indicating when to start filling
587 the queue again. Defaults to the same as iodepth, meaning
588 that fio will attempt to keep the queue full at all times.
589 If iodepth is set to eg 16 and iodepth_low is set to 4, then
590 after fio has filled the queue of 16 requests, it will let
591 the depth drain down to 4 before starting to fill it again.
593 direct=bool If value is true, use non-buffered io. This is usually
594 O_DIRECT. Note that ZFS on Solaris doesn't support direct io.
596 buffered=bool If value is true, use buffered io. This is the opposite
597 of the 'direct' option. Defaults to true.
599 offset=int Start io at the given offset in the file. The data before
600 the given offset will not be touched. This effectively
601 caps the file size at real_size - offset.
603 fsync=int If writing to a file, issue a sync of the dirty data
604 for every number of blocks given. For example, if you give
605 32 as a parameter, fio will sync the file for every 32
606 writes issued. If fio is using non-buffered io, we may
607 not sync the file. The exception is the sg io engine, which
608 synchronizes the disk cache anyway.
610 fdatasync=int Like fsync= but uses fdatasync() to only sync data and not
612 In FreeBSD there is no fdatasync(), this falls back to
615 sync_file_range=str:val Use sync_file_range() for every 'val' number of
616 write operations. Fio will track range of writes that
617 have happened since the last sync_file_range() call. 'str'
618 can currently be one or more of:
620 wait_before SYNC_FILE_RANGE_WAIT_BEFORE
621 write SYNC_FILE_RANGE_WRITE
622 wait_after SYNC_FILE_RANGE_WAIT_AFTER
624 So if you do sync_file_range=wait_before,write:8, fio would
625 use SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE for
626 every 8 writes. Also see the sync_file_range(2) man page.
627 This option is Linux specific.
629 overwrite=bool If true, writes to a file will always overwrite existing
630 data. If the file doesn't already exist, it will be
631 created before the write phase begins. If the file exists
632 and is large enough for the specified write phase, nothing
635 end_fsync=bool If true, fsync file contents when the job exits.
637 fsync_on_close=bool If true, fio will fsync() a dirty file on close.
638 This differs from end_fsync in that it will happen on every
639 file close, not just at the end of the job.
641 rwmixread=int How large a percentage of the mix should be reads.
643 rwmixwrite=int How large a percentage of the mix should be writes. If both
644 rwmixread and rwmixwrite is given and the values do not add
645 up to 100%, the latter of the two will be used to override
646 the first. This may interfere with a given rate setting,
647 if fio is asked to limit reads or writes to a certain rate.
648 If that is the case, then the distribution may be skewed.
650 norandommap Normally fio will cover every block of the file when doing
651 random IO. If this option is given, fio will just get a
652 new random offset without looking at past io history. This
653 means that some blocks may not be read or written, and that
654 some blocks may be read/written more than once. This option
655 is mutually exclusive with verify= if and only if multiple
656 blocksizes (via bsrange=) are used, since fio only tracks
657 complete rewrites of blocks.
659 softrandommap See norandommap. If fio runs with the random block map enabled
660 and it fails to allocate the map, if this option is set it
661 will continue without a random block map. As coverage will
662 not be as complete as with random maps, this option is
665 nice=int Run the job with the given nice value. See man nice(2).
667 prio=int Set the io priority value of this job. Linux limits us to
668 a positive value between 0 and 7, with 0 being the highest.
671 prioclass=int Set the io priority class. See man ionice(1).
673 thinktime=int Stall the job x microseconds after an io has completed before
674 issuing the next. May be used to simulate processing being
675 done by an application. See thinktime_blocks and
679 Only valid if thinktime is set - pretend to spend CPU time
680 doing something with the data received, before falling back
681 to sleeping for the rest of the period specified by
685 Only valid if thinktime is set - control how many blocks
686 to issue, before waiting 'thinktime' usecs. If not set,
687 defaults to 1 which will make fio wait 'thinktime' usecs
690 rate=int Cap the bandwidth used by this job. The number is in bytes/sec,
691 the normal suffix rules apply. You can use rate=500k to limit
692 reads and writes to 500k each, or you can specify read and
693 writes separately. Using rate=1m,500k would limit reads to
694 1MB/sec and writes to 500KB/sec. Capping only reads or
695 writes can be done with rate=,500k or rate=500k,. The former
696 will only limit writes (to 500KB/sec), the latter will only
699 ratemin=int Tell fio to do whatever it can to maintain at least this
700 bandwidth. Failing to meet this requirement, will cause
701 the job to exit. The same format as rate is used for
702 read vs write separation.
704 rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
705 as rate, just specified independently of bandwidth. If the
706 job is given a block size range instead of a fixed value,
707 the smallest block size is used as the metric. The same format
708 as rate is used for read vs write seperation.
710 rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
711 the job to exit. The same format as rate is used for read vs
714 ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
717 cpumask=int Set the CPU affinity of this job. The parameter given is a
718 bitmask of allowed CPU's the job may run on. So if you want
719 the allowed CPUs to be 1 and 5, you would pass the decimal
720 value of (1 << 1 | 1 << 5), or 34. See man
721 sched_setaffinity(2). This may not work on all supported
722 operating systems or kernel versions. This option doesn't
723 work well for a higher CPU count than what you can store in
724 an integer mask, so it can only control cpus 1-32. For
725 boxes with larger CPU counts, use cpus_allowed.
727 cpus_allowed=str Controls the same options as cpumask, but it allows a text
728 setting of the permitted CPUs instead. So to use CPUs 1 and
729 5, you would specify cpus_allowed=1,5. This options also
730 allows a range of CPUs. Say you wanted a binding to CPUs
731 1, 5, and 8-15, you would set cpus_allowed=1,5,8-15.
733 startdelay=time Start this job the specified number of seconds after fio
734 has started. Only useful if the job file contains several
735 jobs, and you want to delay starting some jobs to a certain
738 runtime=time Tell fio to terminate processing after the specified number
739 of seconds. It can be quite hard to determine for how long
740 a specified job will run, so this parameter is handy to
741 cap the total runtime to a given time.
743 time_based If set, fio will run for the duration of the runtime
744 specified even if the file(s) are completely read or
745 written. It will simply loop over the same workload
746 as many times as the runtime allows.
748 ramp_time=time If set, fio will run the specified workload for this amount
749 of time before logging any performance numbers. Useful for
750 letting performance settle before logging results, thus
751 minimizing the runtime required for stable results. Note
752 that the ramp_time is considered lead in time for a job,
753 thus it will increase the total runtime if a special timeout
754 or runtime is specified.
756 invalidate=bool Invalidate the buffer/page cache parts for this file prior
757 to starting io. Defaults to true.
759 sync=bool Use sync io for buffered writes. For the majority of the
760 io engines, this means using O_SYNC.
763 mem=str Fio can use various types of memory as the io unit buffer.
764 The allowed values are:
766 malloc Use memory from malloc(3) as the buffers.
768 shm Use shared memory as the buffers. Allocated
771 shmhuge Same as shm, but use huge pages as backing.
773 mmap Use mmap to allocate buffers. May either be
774 anonymous memory, or can be file backed if
775 a filename is given after the option. The
776 format is mem=mmap:/path/to/file.
778 mmaphuge Use a memory mapped huge file as the buffer
779 backing. Append filename after mmaphuge, ala
780 mem=mmaphuge:/hugetlbfs/file
782 The area allocated is a function of the maximum allowed
783 bs size for the job, multiplied by the io depth given. Note
784 that for shmhuge and mmaphuge to work, the system must have
785 free huge pages allocated. This can normally be checked
786 and set by reading/writing /proc/sys/vm/nr_hugepages on a
787 Linux system. Fio assumes a huge page is 4MB in size. So
788 to calculate the number of huge pages you need for a given
789 job file, add up the io depth of all jobs (normally one unless
790 iodepth= is used) and multiply by the maximum bs set. Then
791 divide that number by the huge page size. You can see the
792 size of the huge pages in /proc/meminfo. If no huge pages
793 are allocated by having a non-zero number in nr_hugepages,
794 using mmaphuge or shmhuge will fail. Also see hugepage-size.
796 mmaphuge also needs to have hugetlbfs mounted and the file
797 location should point there. So if it's mounted in /huge,
798 you would use mem=mmaphuge:/huge/somefile.
800 iomem_align=int This indiciates the memory alignment of the IO memory buffers.
801 Note that the given alignment is applied to the first IO unit
802 buffer, if using iodepth the alignment of the following buffers
803 are given by the bs used. In other words, if using a bs that is
804 a multiple of the page sized in the system, all buffers will
805 be aligned to this value. If using a bs that is not page
806 aligned, the alignment of subsequent IO memory buffers is the
807 sum of the iomem_align and bs used.
810 Defines the size of a huge page. Must at least be equal
811 to the system setting, see /proc/meminfo. Defaults to 4MB.
812 Should probably always be a multiple of megabytes, so using
813 hugepage-size=Xm is the preferred way to set this to avoid
814 setting a non-pow-2 bad value.
816 exitall When one job finishes, terminate the rest. The default is
817 to wait for each job to finish, sometimes that is not the
820 bwavgtime=int Average the calculated bandwidth over the given time. Value
821 is specified in milliseconds.
823 create_serialize=bool If true, serialize the file creating for the jobs.
824 This may be handy to avoid interleaving of data
825 files, which may greatly depend on the filesystem
826 used and even the number of processors in the system.
828 create_fsync=bool fsync the data file after creation. This is the
831 create_on_open=bool Don't pre-setup the files for IO, just create open()
832 when it's time to do IO to that file.
834 pre_read=bool If this is given, files will be pre-read into memory before
835 starting the given IO operation. This will also clear
836 the 'invalidate' flag, since it is pointless to pre-read
837 and then drop the cache. This will only work for IO engines
838 that are seekable, since they allow you to read the same data
839 multiple times. Thus it will not work on eg network or splice
842 unlink=bool Unlink the job files when done. Not the default, as repeated
843 runs of that job would then waste time recreating the file
846 loops=int Run the specified number of iterations of this job. Used
847 to repeat the same workload a given number of times. Defaults
850 do_verify=bool Run the verify phase after a write phase. Only makes sense if
851 verify is set. Defaults to 1.
853 verify=str If writing to a file, fio can verify the file contents
854 after each iteration of the job. The allowed values are:
856 md5 Use an md5 sum of the data area and store
857 it in the header of each block.
859 crc64 Use an experimental crc64 sum of the data
860 area and store it in the header of each
863 crc32c Use a crc32c sum of the data area and store
864 it in the header of each block.
866 crc32c-intel Use hardware assisted crc32c calcuation
867 provided on SSE4.2 enabled processors. Falls
868 back to regular software crc32c, if not
869 supported by the system.
871 crc32 Use a crc32 sum of the data area and store
872 it in the header of each block.
874 crc16 Use a crc16 sum of the data area and store
875 it in the header of each block.
877 crc7 Use a crc7 sum of the data area and store
878 it in the header of each block.
880 sha512 Use sha512 as the checksum function.
882 sha256 Use sha256 as the checksum function.
884 sha1 Use optimized sha1 as the checksum function.
886 meta Write extra information about each io
887 (timestamp, block number etc.). The block
888 number is verified. See also verify_pattern.
890 null Only pretend to verify. Useful for testing
891 internals with ioengine=null, not for much
894 This option can be used for repeated burn-in tests of a
895 system to make sure that the written data is also
896 correctly read back. If the data direction given is
897 a read or random read, fio will assume that it should
898 verify a previously written file. If the data direction
899 includes any form of write, the verify will be of the
902 verifysort=bool If set, fio will sort written verify blocks when it deems
903 it faster to read them back in a sorted manner. This is
904 often the case when overwriting an existing file, since
905 the blocks are already laid out in the file system. You
906 can ignore this option unless doing huge amounts of really
907 fast IO where the red-black tree sorting CPU time becomes
910 verify_offset=int Swap the verification header with data somewhere else
911 in the block before writing. Its swapped back before
914 verify_interval=int Write the verification header at a finer granularity
915 than the blocksize. It will be written for chunks the
916 size of header_interval. blocksize should divide this
919 verify_pattern=str If set, fio will fill the io buffers with this
920 pattern. Fio defaults to filling with totally random
921 bytes, but sometimes it's interesting to fill with a known
922 pattern for io verification purposes. Depending on the
923 width of the pattern, fio will fill 1/2/3/4 bytes of the
924 buffer at the time(it can be either a decimal or a hex number).
925 The verify_pattern if larger than a 32-bit quantity has to
926 be a hex number that starts with either "0x" or "0X". Use
929 verify_fatal=bool Normally fio will keep checking the entire contents
930 before quitting on a block verification failure. If this
931 option is set, fio will exit the job on the first observed
934 verify_dump=bool If set, dump the contents of both the original data
935 block and the data block we read off disk to files. This
936 allows later analysis to inspect just what kind of data
937 corruption occurred. On by default.
939 verify_async=int Fio will normally verify IO inline from the submitting
940 thread. This option takes an integer describing how many
941 async offload threads to create for IO verification instead,
942 causing fio to offload the duty of verifying IO contents
943 to one or more separate threads. If using this offload
944 option, even sync IO engines can benefit from using an
945 iodepth setting higher than 1, as it allows them to have
946 IO in flight while verifies are running.
948 verify_async_cpus=str Tell fio to set the given CPU affinity on the
949 async IO verification threads. See cpus_allowed for the
952 verify_backlog=int Fio will normally verify the written contents of a
953 job that utilizes verify once that job has completed. In
954 other words, everything is written then everything is read
955 back and verified. You may want to verify continually
956 instead for a variety of reasons. Fio stores the meta data
957 associated with an IO block in memory, so for large
958 verify workloads, quite a bit of memory would be used up
959 holding this meta data. If this option is enabled, fio
960 will write only N blocks before verifying these blocks.
962 will verify the previously written blocks before continuing
965 verify_backlog_batch=int Control how many blocks fio will verify
966 if verify_backlog is set. If not set, will default to
967 the value of verify_backlog (meaning the entire queue
968 is read back and verified). If verify_backlog_batch is
969 less than verify_backlog then not all blocks will be verified,
970 if verify_backlog_batch is larger than verify_backlog, some
971 blocks will be verified more than once.
973 stonewall Wait for preceeding jobs in the job file to exit, before
974 starting this one. Can be used to insert serialization
975 points in the job file. A stone wall also implies starting
976 a new reporting group.
978 new_group Start a new reporting group. If this option isn't given,
979 jobs in a file will be part of the same reporting group
980 unless separated by a stone wall (or if it's a group
981 by itself, with the numjobs option).
983 numjobs=int Create the specified number of clones of this job. May be
984 used to setup a larger number of threads/processes doing
985 the same thing. We regard that grouping of jobs as a
988 group_reporting If 'numjobs' is set, it may be interesting to display
989 statistics for the group as a whole instead of for each
990 individual job. This is especially true of 'numjobs' is
991 large, looking at individual thread/process output quickly
992 becomes unwieldy. If 'group_reporting' is specified, fio
993 will show the final report per-group instead of per-job.
995 thread fio defaults to forking jobs, however if this option is
996 given, fio will use pthread_create(3) to create threads
999 zonesize=int Divide a file into zones of the specified size. See zoneskip.
1001 zoneskip=int Skip the specified number of bytes when zonesize data has
1002 been read. The two zone options can be used to only do
1003 io on zones of a file.
1005 write_iolog=str Write the issued io patterns to the specified file. See
1006 read_iolog. Specify a separate file for each job, otherwise
1007 the iologs will be interspersed and the file may be corrupt.
1009 read_iolog=str Open an iolog with the specified file name and replay the
1010 io patterns it contains. This can be used to store a
1011 workload and replay it sometime later. The iolog given
1012 may also be a blktrace binary file, which allows fio
1013 to replay a workload captured by blktrace. See blktrace
1014 for how to capture such logging data. For blktrace replay,
1015 the file needs to be turned into a blkparse binary data
1016 file first (blkparse <device> -o /dev/null -d file_for_fio.bin).
1018 replay_no_stall=int When replaying I/O with read_iolog the default behavior
1019 is to attempt to respect the time stamps within the log and
1020 replay them with the appropriate delay between IOPS. By
1021 setting this variable fio will not respect the timestamps and
1022 attempt to replay them as fast as possible while still
1023 respecting ordering. The result is the same I/O pattern to a
1024 given device, but different timings.
1026 replay_redirect=str While replaying I/O patterns using read_iolog the
1027 default behavior is to replay the IOPS onto the major/minor
1028 device that each IOP was recorded from. This is sometimes
1029 undesireable because on a different machine those major/minor
1030 numbers can map to a different device. Changing hardware on
1031 the same system can also result in a different major/minor
1032 mapping. Replay_redirect causes all IOPS to be replayed onto
1033 the single specified device regardless of the device it was
1034 recorded from. i.e. replay_redirect=/dev/sdc would cause all
1035 IO in the blktrace to be replayed onto /dev/sdc. This means
1036 multiple devices will be replayed onto a single, if the trace
1037 contains multiple devices. If you want multiple devices to be
1038 replayed concurrently to multiple redirected devices you must
1039 blkparse your trace into separate traces and replay them with
1040 independent fio invocations. Unfortuantely this also breaks
1041 the strict time ordering between multiple device accesses.
1043 write_bw_log=str If given, write a bandwidth log of the jobs in this job
1044 file. Can be used to store data of the bandwidth of the
1045 jobs in their lifetime. The included fio_generate_plots
1046 script uses gnuplot to turn these text files into nice
1047 graphs. See write_log_log for behaviour of given
1048 filename. For this option, the postfix is _bw.log.
1050 write_lat_log=str Same as write_bw_log, except that this option stores io
1051 submission, completion, and total latencies instead. If no
1052 filename is given with this option, the default filename of
1053 "jobname_type.log" is used. Even if the filename is given,
1054 fio will still append the type of log. So if one specifies
1058 The actual log names will be foo_slat.log, foo_slat.log,
1059 and foo_lat.log. This helps fio_generate_plot fine the logs
1062 lockmem=int Pin down the specified amount of memory with mlock(2). Can
1063 potentially be used instead of removing memory or booting
1064 with less memory to simulate a smaller amount of memory.
1066 exec_prerun=str Before running this job, issue the command specified
1069 exec_postrun=str After the job completes, issue the command specified
1072 ioscheduler=str Attempt to switch the device hosting the file to the specified
1073 io scheduler before running.
1075 cpuload=int If the job is a CPU cycle eater, attempt to use the specified
1076 percentage of CPU cycles.
1078 cpuchunks=int If the job is a CPU cycle eater, split the load into
1079 cycles of the given time. In microseconds.
1081 disk_util=bool Generate disk utilization statistics, if the platform
1082 supports it. Defaults to on.
1084 disable_lat=bool Disable measurements of total latency numbers. Useful
1085 only for cutting back the number of calls to gettimeofday,
1086 as that does impact performance at really high IOPS rates.
1087 Note that to really get rid of a large amount of these
1088 calls, this option must be used with disable_slat and
1091 disable_clat=bool Disable measurements of completion latency numbers. See
1094 disable_slat=bool Disable measurements of submission latency numbers. See
1097 disable_bw=bool Disable measurements of throughput/bandwidth numbers. See
1100 gtod_reduce=bool Enable all of the gettimeofday() reducing options
1101 (disable_clat, disable_slat, disable_bw) plus reduce
1102 precision of the timeout somewhat to really shrink
1103 the gettimeofday() call count. With this option enabled,
1104 we only do about 0.4% of the gtod() calls we would have
1105 done if all time keeping was enabled.
1107 gtod_cpu=int Sometimes it's cheaper to dedicate a single thread of
1108 execution to just getting the current time. Fio (and
1109 databases, for instance) are very intensive on gettimeofday()
1110 calls. With this option, you can set one CPU aside for
1111 doing nothing but logging current time to a shared memory
1112 location. Then the other threads/processes that run IO
1113 workloads need only copy that segment, instead of entering
1114 the kernel with a gettimeofday() call. The CPU set aside
1115 for doing these time calls will be excluded from other
1116 uses. Fio will manually clear it from the CPU mask of other
1119 continue_on_error=bool Normally fio will exit the job on the first observed
1120 failure. If this option is set, fio will continue the job when
1121 there is a 'non-fatal error' (EIO or EILSEQ) until the runtime
1122 is exceeded or the I/O size specified is completed. If this
1123 option is used, there are two more stats that are appended,
1124 the total error count and the first error. The error field
1125 given in the stats is the first error that was hit during the
1128 cgroup=str Add job to this control group. If it doesn't exist, it will
1129 be created. The system must have a mounted cgroup blkio
1130 mount point for this to work. If your system doesn't have it
1131 mounted, you can do so with:
1133 # mount -t cgroup -o blkio none /cgroup
1135 cgroup_weight=int Set the weight of the cgroup to this value. See
1136 the documentation that comes with the kernel, allowed values
1137 are in the range of 100..1000.
1139 cgroup_nodelete=bool Normally fio will delete the cgroups it has created after
1140 the job completion. To override this behavior and to leave
1141 cgroups around after the job completion, set cgroup_nodelete=1.
1142 This can be useful if one wants to inspect various cgroup
1143 files after job completion. Default: false
1145 uid=int Instead of running as the invoking user, set the user ID to
1146 this value before the thread/process does any work.
1148 gid=int Set group ID, see uid.
1150 6.0 Interpreting the output
1151 ---------------------------
1153 fio spits out a lot of output. While running, fio will display the
1154 status of the jobs created. An example of that would be:
1156 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1158 The characters inside the square brackets denote the current status of
1159 each thread. The possible values (in typical life cycle order) are:
1163 P Thread setup, but not started.
1165 I Thread initialized, waiting.
1166 p Thread running pre-reading file(s).
1167 R Running, doing sequential reads.
1168 r Running, doing random reads.
1169 W Running, doing sequential writes.
1170 w Running, doing random writes.
1171 M Running, doing mixed sequential reads/writes.
1172 m Running, doing mixed random reads/writes.
1173 F Running, currently waiting for fsync()
1174 V Running, doing verification of written data.
1175 E Thread exited, not reaped by main thread yet.
1178 The other values are fairly self explanatory - number of threads
1179 currently running and doing io, rate of io since last check (read speed
1180 listed first, then write speed), and the estimated completion percentage
1181 and time for the running group. It's impossible to estimate runtime of
1182 the following groups (if any).
1184 When fio is done (or interrupted by ctrl-c), it will show the data for
1185 each thread, group of threads, and disks in that order. For each data
1186 direction, the output looks like:
1188 Client1 (g=0): err= 0:
1189 write: io= 32MB, bw= 666KB/s, runt= 50320msec
1190 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
1191 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
1192 bw (KB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
1193 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
1194 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
1195 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1196 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1197 issued r/w: total=0/32768, short=0/0
1198 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
1199 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
1201 The client number is printed, along with the group id and error of that
1202 thread. Below is the io statistics, here for writes. In the order listed,
1205 io= Number of megabytes io performed
1206 bw= Average bandwidth rate
1207 runt= The runtime of that thread
1208 slat= Submission latency (avg being the average, stdev being the
1209 standard deviation). This is the time it took to submit
1210 the io. For sync io, the slat is really the completion
1211 latency, since queue/complete is one operation there. This
1212 value can be in milliseconds or microseconds, fio will choose
1213 the most appropriate base and print that. In the example
1214 above, milliseconds is the best scale.
1215 clat= Completion latency. Same names as slat, this denotes the
1216 time from submission to completion of the io pieces. For
1217 sync io, clat will usually be equal (or very close) to 0,
1218 as the time from submit to complete is basically just
1219 CPU time (io has already been done, see slat explanation).
1220 bw= Bandwidth. Same names as the xlat stats, but also includes
1221 an approximate percentage of total aggregate bandwidth
1222 this thread received in this group. This last value is
1223 only really useful if the threads in this group are on the
1224 same disk, since they are then competing for disk access.
1225 cpu= CPU usage. User and system time, along with the number
1226 of context switches this thread went through, usage of
1227 system and user time, and finally the number of major
1228 and minor page faults.
1229 IO depths= The distribution of io depths over the job life time. The
1230 numbers are divided into powers of 2, so for example the
1231 16= entries includes depths up to that value but higher
1232 than the previous entry. In other words, it covers the
1233 range from 16 to 31.
1234 IO submit= How many pieces of IO were submitting in a single submit
1235 call. Each entry denotes that amount and below, until
1236 the previous entry - eg, 8=100% mean that we submitted
1237 anywhere in between 5-8 ios per submit call.
1238 IO complete= Like the above submit number, but for completions instead.
1239 IO issued= The number of read/write requests issued, and how many
1241 IO latencies= The distribution of IO completion latencies. This is the
1242 time from when IO leaves fio and when it gets completed.
1243 The numbers follow the same pattern as the IO depths,
1244 meaning that 2=1.6% means that 1.6% of the IO completed
1245 within 2 msecs, 20=12.8% means that 12.8% of the IO
1246 took more than 10 msecs, but less than (or equal to) 20 msecs.
1248 After each client has been listed, the group statistics are printed. They
1249 will look like this:
1251 Run status group 0 (all jobs):
1252 READ: io=64MB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
1253 WRITE: io=64MB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
1255 For each data direction, it prints:
1257 io= Number of megabytes io performed.
1258 aggrb= Aggregate bandwidth of threads in this group.
1259 minb= The minimum average bandwidth a thread saw.
1260 maxb= The maximum average bandwidth a thread saw.
1261 mint= The smallest runtime of the threads in that group.
1262 maxt= The longest runtime of the threads in that group.
1264 And finally, the disk statistics are printed. They will look like this:
1266 Disk stats (read/write):
1267 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
1269 Each value is printed for both reads and writes, with reads first. The
1272 ios= Number of ios performed by all groups.
1273 merge= Number of merges io the io scheduler.
1274 ticks= Number of ticks we kept the disk busy.
1275 io_queue= Total time spent in the disk queue.
1276 util= The disk utilization. A value of 100% means we kept the disk
1277 busy constantly, 50% would be a disk idling half of the time.
1283 For scripted usage where you typically want to generate tables or graphs
1284 of the results, fio can output the results in a semicolon separated format.
1285 The format is one long line of values, such as:
1287 2;card0;0;0;7139336;121836;60004;1;10109;27.932460;116.933948;220;126861;3495.446807;1085.368601;226;126864;3523.635629;1089.012448;24063;99944;50.275485%;59818.274627;5540.657370;7155060;122104;60004;1;8338;29.086342;117.839068;388;128077;5032.488518;1234.785715;391;128085;5061.839412;1236.909129;23436;100928;50.287926%;59964.832030;5644.844189;14.595833%;19.394167%;123706;0;7313;0.1%;0.1%;0.1%;0.1%;0.1%;0.1%;100.0%;0.00%;0.00%;0.00%;0.00%;0.00%;0.00%;0.01%;0.02%;0.05%;0.16%;6.04%;40.40%;52.68%;0.64%;0.01%;0.00%;0.01%;0.00%;0.00%;0.00%;0.00%;0.00%
1288 A description of this job goes here.
1290 The job description (if provided) follows on a second line.
1292 To enable terse output, use the --minimal command line option. The first
1293 value is the version of the terse output format. If the output has to
1294 be changed for some reason, this number will be incremented by 1 to
1295 signify that change.
1297 Split up, the format is as follows:
1299 version, jobname, groupid, error
1301 KB IO, bandwidth (KB/sec), runtime (msec)
1302 Submission latency: min, max, mean, deviation
1303 Completion latency: min, max, mean, deviation
1304 Total latency: min, max, mean, deviation
1305 Bw: min, max, aggregate percentage of total, mean, deviation
1307 KB IO, bandwidth (KB/sec), runtime (msec)
1308 Submission latency: min, max, mean, deviation
1309 Completion latency: min, max, mean, deviation
1310 Total latency: min, max, mean, deviation
1311 Bw: min, max, aggregate percentage of total, mean, deviation
1312 CPU usage: user, system, context switches, major faults, minor faults
1313 IO depths: <=1, 2, 4, 8, 16, 32, >=64
1314 IO latencies microseconds: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1315 IO latencies milliseconds: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1316 Additional Info (dependant on continue_on_error, default off): total # errors, first error code
1318 Additional Info (dependant on description being set): Text description