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. So if you want to specify 4096, you could either write
230 out '4096' or just give 4k. The suffixes signify base 2 values, so
231 1024 is 1k and 1024k is 1m and so on. If the option accepts an upper
232 and lower range, use a colon ':' or minus '-' to separate such values.
233 May also include a prefix to indicate numbers base. If 0x is used,
234 the number is assumed to be hexadecimal. See irange.
235 bool Boolean. Usually parsed as an integer, however only defined for
236 true and false (1 and 0).
237 irange Integer range with suffix. Allows value range to be given, such
238 as 1024-4096. A colon may also be used as the separator, eg
239 1k:4k. If the option allows two sets of ranges, they can be
240 specified with a ',' or '/' delimiter: 1k-4k/8k-32k. Also see
243 With the above in mind, here follows the complete list of fio job
246 name=str ASCII name of the job. This may be used to override the
247 name printed by fio for this job. Otherwise the job
248 name is used. On the command line this parameter has the
249 special purpose of also signaling the start of a new
252 description=str Text description of the job. Doesn't do anything except
253 dump this text description when this job is run. It's
256 directory=str Prefix filenames with this directory. Used to place files
257 in a different location than "./".
259 filename=str Fio normally makes up a filename based on the job name,
260 thread number, and file number. If you want to share
261 files between threads in a job or several jobs, specify
262 a filename for each of them to override the default. If
263 the ioengine used is 'net', the filename is the host, port,
264 and protocol to use in the format of =host/port/protocol.
265 See ioengine=net for more. If the ioengine is file based, you
266 can specify a number of files by separating the names with a
267 ':' colon. So if you wanted a job to open /dev/sda and /dev/sdb
268 as the two working files, you would use
269 filename=/dev/sda:/dev/sdb. If the wanted filename does need to
270 include a colon, then escape that with a '\' character. For
271 instance, if the filename is "/dev/dsk/foo@3,0:c", then you would
272 use filename="/dev/dsk/foo@3,0\:c". '-' is a reserved name,
273 meaning stdin or stdout. Which of the two depends on the read/write
276 opendir=str Tell fio to recursively add any file it can find in this
277 directory and down the file system tree.
279 lockfile=str Fio defaults to not locking any files before it does
280 IO to them. If a file or file descriptor is shared, fio
281 can serialize IO to that file to make the end result
282 consistent. This is usual for emulating real workloads that
283 share files. The lock modes are:
285 none No locking. The default.
286 exclusive Only one thread/process may do IO,
287 excluding all others.
288 readwrite Read-write locking on the file. Many
289 readers may access the file at the
290 same time, but writes get exclusive
293 The option may be post-fixed with a lock batch number. If
294 set, then each thread/process may do that amount of IOs to
295 the file before giving up the lock. Since lock acquisition is
296 expensive, batching the lock/unlocks will speed up IO.
299 rw=str Type of io pattern. Accepted values are:
301 read Sequential reads
302 write Sequential writes
303 randwrite Random writes
304 randread Random reads
305 rw Sequential mixed reads and writes
306 randrw Random mixed reads and writes
308 For the mixed io types, the default is to split them 50/50.
309 For certain types of io the result may still be skewed a bit,
310 since the speed may be different. It is possible to specify
311 a number of IO's to do before getting a new offset - this
312 is only useful for random IO, where fio would normally
313 generate a new random offset for every IO. If you append
314 eg 8 to randread, you would get a new random offset for
315 every 8 IO's. The result would be a seek for only every 8
316 IO's, instead of for every IO. Use rw=randread:8 to specify
319 kb_base=int The base unit for a kilobyte. The defacto base is 2^10, 1024.
320 Storage manufacturers like to use 10^3 or 1000 as a base
321 ten unit instead, for obvious reasons. Allow values are
322 1024 or 1000, with 1024 being the default.
324 randrepeat=bool For random IO workloads, seed the generator in a predictable
325 way so that results are repeatable across repetitions.
327 fallocate=bool By default, fio will use fallocate() to advise the system
328 of the size of the file we are going to write. This can be
329 turned off with fallocate=0. May not be available on all
332 fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
333 on what IO patterns it is likely to issue. Sometimes you
334 want to test specific IO patterns without telling the
335 kernel about it, in which case you can disable this option.
336 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
337 IO and POSIX_FADV_RANDOM for random IO.
339 size=int The total size of file io for this job. Fio will run until
340 this many bytes has been transferred, unless runtime is
341 limited by other options (such as 'runtime', for instance).
342 Unless specific nrfiles and filesize options are given,
343 fio will divide this size between the available files
344 specified by the job.
346 filesize=int Individual file sizes. May be a range, in which case fio
347 will select sizes for files at random within the given range
348 and limited to 'size' in total (if that is given). If not
349 given, each created file is the same size.
351 fill_device=bool Sets size to something really large and waits for ENOSPC (no
352 space left on device) as the terminating condition. Only makes
353 sense with sequential write. For a read workload, the mount
354 point will be filled first then IO started on the result.
357 bs=int The block size used for the io units. Defaults to 4k. Values
358 can be given for both read and writes. If a single int is
359 given, it will apply to both. If a second int is specified
360 after a comma, it will apply to writes only. In other words,
361 the format is either bs=read_and_write or bs=read,write.
362 bs=4k,8k will thus use 4k blocks for reads, and 8k blocks
363 for writes. If you only wish to set the write size, you
364 can do so by passing an empty read size - bs=,8k will set
365 8k for writes and leave the read default value.
368 ba=int At what boundary to align random IO offsets. Defaults to
369 the same as 'blocksize' the minimum blocksize given.
370 Minimum alignment is typically 512b for using direct IO,
371 though it usually depends on the hardware block size. This
372 option is mutually exclusive with using a random map for
373 files, so it will turn off that option.
375 blocksize_range=irange
376 bsrange=irange Instead of giving a single block size, specify a range
377 and fio will mix the issued io block sizes. The issued
378 io unit will always be a multiple of the minimum value
379 given (also see bs_unaligned). Applies to both reads and
380 writes, however a second range can be given after a comma.
383 bssplit=str Sometimes you want even finer grained control of the
384 block sizes issued, not just an even split between them.
385 This option allows you to weight various block sizes,
386 so that you are able to define a specific amount of
387 block sizes issued. The format for this option is:
389 bssplit=blocksize/percentage:blocksize/percentage
391 for as many block sizes as needed. So if you want to define
392 a workload that has 50% 64k blocks, 10% 4k blocks, and
393 40% 32k blocks, you would write:
395 bssplit=4k/10:64k/50:32k/40
397 Ordering does not matter. If the percentage is left blank,
398 fio will fill in the remaining values evenly. So a bssplit
399 option like this one:
401 bssplit=4k/50:1k/:32k/
403 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
404 always add up to 100, if bssplit is given a range that adds
405 up to more, it will error out.
407 bssplit also supports giving separate splits to reads and
408 writes. The format is identical to what bs= accepts. You
409 have to separate the read and write parts with a comma. So
410 if you want a workload that has 50% 2k reads and 50% 4k reads,
411 while having 90% 4k writes and 10% 8k writes, you would
414 bssplit=2k/50:4k/50,4k/90,8k/10
417 bs_unaligned If this option is given, any byte size value within bsrange
418 may be used as a block range. This typically wont work with
419 direct IO, as that normally requires sector alignment.
421 zero_buffers If this option is given, fio will init the IO buffers to
422 all zeroes. The default is to fill them with random data.
424 refill_buffers If this option is given, fio will refill the IO buffers
425 on every submit. The default is to only fill it at init
426 time and reuse that data. Only makes sense if zero_buffers
427 isn't specified, naturally. If data verification is enabled,
428 refill_buffers is also automatically enabled.
430 nrfiles=int Number of files to use for this job. Defaults to 1.
432 openfiles=int Number of files to keep open at the same time. Defaults to
433 the same as nrfiles, can be set smaller to limit the number
436 file_service_type=str Defines how fio decides which file from a job to
437 service next. The following types are defined:
439 random Just choose a file at random.
441 roundrobin Round robin over open files. This
444 sequential Finish one file before moving on to
445 the next. Multiple files can still be
446 open depending on 'openfiles'.
448 The string can have a number appended, indicating how
449 often to switch to a new file. So if option random:4 is
450 given, fio will switch to a new random file after 4 ios
453 ioengine=str Defines how the job issues io to the file. The following
456 sync Basic read(2) or write(2) io. lseek(2) is
457 used to position the io location.
459 psync Basic pread(2) or pwrite(2) io.
461 vsync Basic readv(2) or writev(2) IO.
463 libaio Linux native asynchronous io. Note that Linux
464 may only support queued behaviour with
465 non-buffered IO (set direct=1 or buffered=0).
467 posixaio glibc posix asynchronous io.
469 solarisaio Solaris native asynchronous io.
471 mmap File is memory mapped and data copied
472 to/from using memcpy(3).
474 splice splice(2) is used to transfer the data and
475 vmsplice(2) to transfer data from user
478 syslet-rw Use the syslet system calls to make
479 regular read/write async.
481 sg SCSI generic sg v3 io. May either be
482 synchronous using the SG_IO ioctl, or if
483 the target is an sg character device
484 we use read(2) and write(2) for asynchronous
487 null Doesn't transfer any data, just pretends
488 to. This is mainly used to exercise fio
489 itself and for debugging/testing purposes.
491 net Transfer over the network to given host:port.
492 'filename' must be set appropriately to
493 filename=host/port/protocol regardless of send
494 or receive, if the latter only the port
495 argument is used. 'host' may be an IP address
496 or hostname, port is the port number to be used,
497 and protocol may be 'udp' or 'tcp'. If no
498 protocol is given, TCP is used.
500 netsplice Like net, but uses splice/vmsplice to
501 map data and send/receive.
503 cpuio Doesn't transfer any data, but burns CPU
504 cycles according to the cpuload= and
505 cpucycle= options. Setting cpuload=85
506 will cause that job to do nothing but burn
507 85% of the CPU. In case of SMP machines,
508 use numjobs=<no_of_cpu> to get desired CPU
509 usage, as the cpuload only loads a single
510 CPU at the desired rate.
512 guasi The GUASI IO engine is the Generic Userspace
513 Asyncronous Syscall Interface approach
516 http://www.xmailserver.org/guasi-lib.html
518 for more info on GUASI.
520 external Prefix to specify loading an external
521 IO engine object file. Append the engine
522 filename, eg ioengine=external:/tmp/foo.o
523 to load ioengine foo.o in /tmp.
525 iodepth=int This defines how many io units to keep in flight against
526 the file. The default is 1 for each file defined in this
527 job, can be overridden with a larger value for higher
530 iodepth_batch_submit=int
531 iodepth_batch=int This defines how many pieces of IO to submit at once.
532 It defaults to 1 which means that we submit each IO
533 as soon as it is available, but can be raised to submit
534 bigger batches of IO at the time.
536 iodepth_batch_complete=int This defines how many pieces of IO to retrieve
537 at once. It defaults to 1 which means that we'll ask
538 for a minimum of 1 IO in the retrieval process from
539 the kernel. The IO retrieval will go on until we
540 hit the limit set by iodepth_low. If this variable is
541 set to 0, then fio will always check for completed
542 events before queuing more IO. This helps reduce
543 IO latency, at the cost of more retrieval system calls.
545 iodepth_low=int The low water mark indicating when to start filling
546 the queue again. Defaults to the same as iodepth, meaning
547 that fio will attempt to keep the queue full at all times.
548 If iodepth is set to eg 16 and iodepth_low is set to 4, then
549 after fio has filled the queue of 16 requests, it will let
550 the depth drain down to 4 before starting to fill it again.
552 direct=bool If value is true, use non-buffered io. This is usually
555 buffered=bool If value is true, use buffered io. This is the opposite
556 of the 'direct' option. Defaults to true.
558 offset=int Start io at the given offset in the file. The data before
559 the given offset will not be touched. This effectively
560 caps the file size at real_size - offset.
562 fsync=int If writing to a file, issue a sync of the dirty data
563 for every number of blocks given. For example, if you give
564 32 as a parameter, fio will sync the file for every 32
565 writes issued. If fio is using non-buffered io, we may
566 not sync the file. The exception is the sg io engine, which
567 synchronizes the disk cache anyway.
569 fdatasync=int Like fsync= but uses fdatasync() to only sync data and not
571 In FreeBSD there is no fdatasync(), this falls back to
574 sync_file_range=str:val Use sync_file_range() for every 'val' number of
575 write operations. Fio will track range of writes that
576 have happened since the last sync_file_range() call. 'str'
577 can currently be one or more of:
579 wait_before SYNC_FILE_RANGE_WAIT_BEFORE
580 write SYNC_FILE_RANGE_WRITE
581 wait_after SYNC_FILE_RANGE_WAIT_AFTER
583 So if you do sync_file_range=wait_before,write:8, fio would
584 use SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE for
585 every 8 writes. Also see the sync_file_range(2) man page.
586 This option is Linux specific.
588 overwrite=bool If true, writes to a file will always overwrite existing
589 data. If the file doesn't already exist, it will be
590 created before the write phase begins. If the file exists
591 and is large enough for the specified write phase, nothing
594 end_fsync=bool If true, fsync file contents when the job exits.
596 fsync_on_close=bool If true, fio will fsync() a dirty file on close.
597 This differs from end_fsync in that it will happen on every
598 file close, not just at the end of the job.
600 rwmixread=int How large a percentage of the mix should be reads.
602 rwmixwrite=int How large a percentage of the mix should be writes. If both
603 rwmixread and rwmixwrite is given and the values do not add
604 up to 100%, the latter of the two will be used to override
605 the first. This may interfere with a given rate setting,
606 if fio is asked to limit reads or writes to a certain rate.
607 If that is the case, then the distribution may be skewed.
609 norandommap Normally fio will cover every block of the file when doing
610 random IO. If this option is given, fio will just get a
611 new random offset without looking at past io history. This
612 means that some blocks may not be read or written, and that
613 some blocks may be read/written more than once. This option
614 is mutually exclusive with verify= if and only if multiple
615 blocksizes (via bsrange=) are used, since fio only tracks
616 complete rewrites of blocks.
618 softrandommap See norandommap. If fio runs with the random block map enabled
619 and it fails to allocate the map, if this option is set it
620 will continue without a random block map. As coverage will
621 not be as complete as with random maps, this option is
624 nice=int Run the job with the given nice value. See man nice(2).
626 prio=int Set the io priority value of this job. Linux limits us to
627 a positive value between 0 and 7, with 0 being the highest.
630 prioclass=int Set the io priority class. See man ionice(1).
632 thinktime=int Stall the job x microseconds after an io has completed before
633 issuing the next. May be used to simulate processing being
634 done by an application. See thinktime_blocks and
638 Only valid if thinktime is set - pretend to spend CPU time
639 doing something with the data received, before falling back
640 to sleeping for the rest of the period specified by
644 Only valid if thinktime is set - control how many blocks
645 to issue, before waiting 'thinktime' usecs. If not set,
646 defaults to 1 which will make fio wait 'thinktime' usecs
649 rate=int Cap the bandwidth used by this job. The number is in bytes/sec,
650 the normal suffix rules apply. You can use rate=500k to limit
651 reads and writes to 500k each, or you can specify read and
652 writes separately. Using rate=1m,500k would limit reads to
653 1MB/sec and writes to 500KB/sec. Capping only reads or
654 writes can be done with rate=,500k or rate=500k,. The former
655 will only limit writes (to 500KB/sec), the latter will only
658 ratemin=int Tell fio to do whatever it can to maintain at least this
659 bandwidth. Failing to meet this requirement, will cause
660 the job to exit. The same format as rate is used for
661 read vs write separation.
663 rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
664 as rate, just specified independently of bandwidth. If the
665 job is given a block size range instead of a fixed value,
666 the smallest block size is used as the metric. The same format
667 as rate is used for read vs write seperation.
669 rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
670 the job to exit. The same format as rate is used for read vs
673 ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
676 cpumask=int Set the CPU affinity of this job. The parameter given is a
677 bitmask of allowed CPU's the job may run on. So if you want
678 the allowed CPUs to be 1 and 5, you would pass the decimal
679 value of (1 << 1 | 1 << 5), or 34. See man
680 sched_setaffinity(2). This may not work on all supported
681 operating systems or kernel versions. This option doesn't
682 work well for a higher CPU count than what you can store in
683 an integer mask, so it can only control cpus 1-32. For
684 boxes with larger CPU counts, use cpus_allowed.
686 cpus_allowed=str Controls the same options as cpumask, but it allows a text
687 setting of the permitted CPUs instead. So to use CPUs 1 and
688 5, you would specify cpus_allowed=1,5. This options also
689 allows a range of CPUs. Say you wanted a binding to CPUs
690 1, 5, and 8-15, you would set cpus_allowed=1,5,8-15.
692 startdelay=time Start this job the specified number of seconds after fio
693 has started. Only useful if the job file contains several
694 jobs, and you want to delay starting some jobs to a certain
697 runtime=time Tell fio to terminate processing after the specified number
698 of seconds. It can be quite hard to determine for how long
699 a specified job will run, so this parameter is handy to
700 cap the total runtime to a given time.
702 time_based If set, fio will run for the duration of the runtime
703 specified even if the file(s) are completely read or
704 written. It will simply loop over the same workload
705 as many times as the runtime allows.
707 ramp_time=time If set, fio will run the specified workload for this amount
708 of time before logging any performance numbers. Useful for
709 letting performance settle before logging results, thus
710 minimizing the runtime required for stable results. Note
711 that the ramp_time is considered lead in time for a job,
712 thus it will increase the total runtime if a special timeout
713 or runtime is specified.
715 invalidate=bool Invalidate the buffer/page cache parts for this file prior
716 to starting io. Defaults to true.
718 sync=bool Use sync io for buffered writes. For the majority of the
719 io engines, this means using O_SYNC.
722 mem=str Fio can use various types of memory as the io unit buffer.
723 The allowed values are:
725 malloc Use memory from malloc(3) as the buffers.
727 shm Use shared memory as the buffers. Allocated
730 shmhuge Same as shm, but use huge pages as backing.
732 mmap Use mmap to allocate buffers. May either be
733 anonymous memory, or can be file backed if
734 a filename is given after the option. The
735 format is mem=mmap:/path/to/file.
737 mmaphuge Use a memory mapped huge file as the buffer
738 backing. Append filename after mmaphuge, ala
739 mem=mmaphuge:/hugetlbfs/file
741 The area allocated is a function of the maximum allowed
742 bs size for the job, multiplied by the io depth given. Note
743 that for shmhuge and mmaphuge to work, the system must have
744 free huge pages allocated. This can normally be checked
745 and set by reading/writing /proc/sys/vm/nr_hugepages on a
746 Linux system. Fio assumes a huge page is 4MB in size. So
747 to calculate the number of huge pages you need for a given
748 job file, add up the io depth of all jobs (normally one unless
749 iodepth= is used) and multiply by the maximum bs set. Then
750 divide that number by the huge page size. You can see the
751 size of the huge pages in /proc/meminfo. If no huge pages
752 are allocated by having a non-zero number in nr_hugepages,
753 using mmaphuge or shmhuge will fail. Also see hugepage-size.
755 mmaphuge also needs to have hugetlbfs mounted and the file
756 location should point there. So if it's mounted in /huge,
757 you would use mem=mmaphuge:/huge/somefile.
759 iomem_align=int This indiciates the memory alignment of the IO memory buffers.
760 Note that the given alignment is applied to the first IO unit
761 buffer, if using iodepth the alignment of the following buffers
762 are given by the bs used. In other words, if using a bs that is
763 a multiple of the page sized in the system, all buffers will
764 be aligned to this value. If using a bs that is not page
765 aligned, the alignment of subsequent IO memory buffers is the
766 sum of the iomem_align and bs used.
769 Defines the size of a huge page. Must at least be equal
770 to the system setting, see /proc/meminfo. Defaults to 4MB.
771 Should probably always be a multiple of megabytes, so using
772 hugepage-size=Xm is the preferred way to set this to avoid
773 setting a non-pow-2 bad value.
775 exitall When one job finishes, terminate the rest. The default is
776 to wait for each job to finish, sometimes that is not the
779 bwavgtime=int Average the calculated bandwidth over the given time. Value
780 is specified in milliseconds.
782 create_serialize=bool If true, serialize the file creating for the jobs.
783 This may be handy to avoid interleaving of data
784 files, which may greatly depend on the filesystem
785 used and even the number of processors in the system.
787 create_fsync=bool fsync the data file after creation. This is the
790 create_on_open=bool Don't pre-setup the files for IO, just create open()
791 when it's time to do IO to that file.
793 pre_read=bool If this is given, files will be pre-read into memory before
794 starting the given IO operation. This will also clear
795 the 'invalidate' flag, since it is pointless to pre-read
796 and then drop the cache. This will only work for IO engines
797 that are seekable, since they allow you to read the same data
798 multiple times. Thus it will not work on eg network or splice
801 unlink=bool Unlink the job files when done. Not the default, as repeated
802 runs of that job would then waste time recreating the file
805 loops=int Run the specified number of iterations of this job. Used
806 to repeat the same workload a given number of times. Defaults
809 do_verify=bool Run the verify phase after a write phase. Only makes sense if
810 verify is set. Defaults to 1.
812 verify=str If writing to a file, fio can verify the file contents
813 after each iteration of the job. The allowed values are:
815 md5 Use an md5 sum of the data area and store
816 it in the header of each block.
818 crc64 Use an experimental crc64 sum of the data
819 area and store it in the header of each
822 crc32c Use a crc32c sum of the data area and store
823 it in the header of each block.
825 crc32c-intel Use hardware assisted crc32c calcuation
826 provided on SSE4.2 enabled processors. Falls
827 back to regular software crc32c, if not
828 supported by the system.
830 crc32 Use a crc32 sum of the data area and store
831 it in the header of each block.
833 crc16 Use a crc16 sum of the data area and store
834 it in the header of each block.
836 crc7 Use a crc7 sum of the data area and store
837 it in the header of each block.
839 sha512 Use sha512 as the checksum function.
841 sha256 Use sha256 as the checksum function.
843 sha1 Use optimized sha1 as the checksum function.
845 meta Write extra information about each io
846 (timestamp, block number etc.). The block
849 null Only pretend to verify. Useful for testing
850 internals with ioengine=null, not for much
853 This option can be used for repeated burn-in tests of a
854 system to make sure that the written data is also
855 correctly read back. If the data direction given is
856 a read or random read, fio will assume that it should
857 verify a previously written file. If the data direction
858 includes any form of write, the verify will be of the
861 verifysort=bool If set, fio will sort written verify blocks when it deems
862 it faster to read them back in a sorted manner. This is
863 often the case when overwriting an existing file, since
864 the blocks are already laid out in the file system. You
865 can ignore this option unless doing huge amounts of really
866 fast IO where the red-black tree sorting CPU time becomes
869 verify_offset=int Swap the verification header with data somewhere else
870 in the block before writing. Its swapped back before
873 verify_interval=int Write the verification header at a finer granularity
874 than the blocksize. It will be written for chunks the
875 size of header_interval. blocksize should divide this
878 verify_pattern=str If set, fio will fill the io buffers with this
879 pattern. Fio defaults to filling with totally random
880 bytes, but sometimes it's interesting to fill with a known
881 pattern for io verification purposes. Depending on the
882 width of the pattern, fio will fill 1/2/3/4 bytes of the
883 buffer at the time(it can be either a decimal or a hex number).
884 The verify_pattern if larger than a 32-bit quantity has to
885 be a hex number that starts with either "0x" or "0X".
887 verify_fatal=bool Normally fio will keep checking the entire contents
888 before quitting on a block verification failure. If this
889 option is set, fio will exit the job on the first observed
892 verify_async=int Fio will normally verify IO inline from the submitting
893 thread. This option takes an integer describing how many
894 async offload threads to create for IO verification instead,
895 causing fio to offload the duty of verifying IO contents
896 to one or more separate threads. If using this offload
897 option, even sync IO engines can benefit from using an
898 iodepth setting higher than 1, as it allows them to have
899 IO in flight while verifies are running.
901 verify_async_cpus=str Tell fio to set the given CPU affinity on the
902 async IO verification threads. See cpus_allowed for the
905 verify_backlog=int Fio will normally verify the written contents of a
906 job that utilizes verify once that job has completed. In
907 other words, everything is written then everything is read
908 back and verified. You may want to verify continually
909 instead for a variety of reasons. Fio stores the meta data
910 associated with an IO block in memory, so for large
911 verify workloads, quite a bit of memory would be used up
912 holding this meta data. If this option is enabled, fio
913 will verify the previously written blocks before continuing
916 verify_backlog_batch=int Control how many blocks fio will verify
917 if verify_backlog is set. If not set, will default to
918 the value of verify_backlog (meaning the entire queue
919 is read back and verified).
921 stonewall Wait for preceeding jobs in the job file to exit, before
922 starting this one. Can be used to insert serialization
923 points in the job file. A stone wall also implies starting
924 a new reporting group.
926 new_group Start a new reporting group. If this option isn't given,
927 jobs in a file will be part of the same reporting group
928 unless separated by a stone wall (or if it's a group
929 by itself, with the numjobs option).
931 numjobs=int Create the specified number of clones of this job. May be
932 used to setup a larger number of threads/processes doing
933 the same thing. We regard that grouping of jobs as a
936 group_reporting If 'numjobs' is set, it may be interesting to display
937 statistics for the group as a whole instead of for each
938 individual job. This is especially true of 'numjobs' is
939 large, looking at individual thread/process output quickly
940 becomes unwieldy. If 'group_reporting' is specified, fio
941 will show the final report per-group instead of per-job.
943 thread fio defaults to forking jobs, however if this option is
944 given, fio will use pthread_create(3) to create threads
947 zonesize=int Divide a file into zones of the specified size. See zoneskip.
949 zoneskip=int Skip the specified number of bytes when zonesize data has
950 been read. The two zone options can be used to only do
951 io on zones of a file.
953 write_iolog=str Write the issued io patterns to the specified file. See
956 read_iolog=str Open an iolog with the specified file name and replay the
957 io patterns it contains. This can be used to store a
958 workload and replay it sometime later. The iolog given
959 may also be a blktrace binary file, which allows fio
960 to replay a workload captured by blktrace. See blktrace
961 for how to capture such logging data. For blktrace replay,
962 the file needs to be turned into a blkparse binary data
963 file first (blkparse <device> -o /dev/null -d file_for_fio.bin).
965 write_bw_log=str If given, write a bandwidth log of the jobs in this job
966 file. Can be used to store data of the bandwidth of the
967 jobs in their lifetime. The included fio_generate_plots
968 script uses gnuplot to turn these text files into nice
969 graphs. See write_log_log for behaviour of given
970 filename. For this option, the postfix is _bw.log.
972 write_lat_log=str Same as write_bw_log, except that this option stores io
973 completion latencies instead. If no filename is given
974 with this option, the default filename of "jobname_type.log"
975 is used. Even if the filename is given, fio will still
976 append the type of log. So if one specifies
980 The actual log names will be foo_clat.log and foo_slat.log.
981 This helps fio_generate_plot fine the logs automatically.
983 lockmem=int Pin down the specified amount of memory with mlock(2). Can
984 potentially be used instead of removing memory or booting
985 with less memory to simulate a smaller amount of memory.
987 exec_prerun=str Before running this job, issue the command specified
990 exec_postrun=str After the job completes, issue the command specified
993 ioscheduler=str Attempt to switch the device hosting the file to the specified
994 io scheduler before running.
996 cpuload=int If the job is a CPU cycle eater, attempt to use the specified
997 percentage of CPU cycles.
999 cpuchunks=int If the job is a CPU cycle eater, split the load into
1000 cycles of the given time. In microseconds.
1002 disk_util=bool Generate disk utilization statistics, if the platform
1003 supports it. Defaults to on.
1005 disable_clat=bool Disable measurements of completion latency numbers. Useful
1006 only for cutting back the number of calls to gettimeofday,
1007 as that does impact performance at really high IOPS rates.
1008 Note that to really get rid of a large amount of these
1009 calls, this option must be used with disable_slat and
1012 disable_slat=bool Disable measurements of submission latency numbers. See
1015 disable_bw=bool Disable measurements of throughput/bandwidth numbers. See
1018 gtod_reduce=bool Enable all of the gettimeofday() reducing options
1019 (disable_clat, disable_slat, disable_bw) plus reduce
1020 precision of the timeout somewhat to really shrink
1021 the gettimeofday() call count. With this option enabled,
1022 we only do about 0.4% of the gtod() calls we would have
1023 done if all time keeping was enabled.
1025 gtod_cpu=int Sometimes it's cheaper to dedicate a single thread of
1026 execution to just getting the current time. Fio (and
1027 databases, for instance) are very intensive on gettimeofday()
1028 calls. With this option, you can set one CPU aside for
1029 doing nothing but logging current time to a shared memory
1030 location. Then the other threads/processes that run IO
1031 workloads need only copy that segment, instead of entering
1032 the kernel with a gettimeofday() call. The CPU set aside
1033 for doing these time calls will be excluded from other
1034 uses. Fio will manually clear it from the CPU mask of other
1037 continue_on_error=bool Normally fio will exit the job on the first observed
1038 failure. If this option is set, fio will continue the job when
1039 there is a 'non-fatal error' (EIO or EILSEQ) until the runtime
1040 is exceeded or the I/O size specified is completed. If this
1041 option is used, there are two more stats that are appended,
1042 the total error count and the first error. The error field
1043 given in the stats is the first error that was hit during the
1046 cgroup=str Add job to this control group. If it doesn't exist, it will
1047 be created. The system must have a mounted cgroup blkio
1048 mount point for this to work. If your system doesn't have it
1049 mounted, you can do so with:
1051 # mount -t cgroup -o blkio none /cgroup
1053 cgroup_weight=int Set the weight of the cgroup to this value. See
1054 the documentation that comes with the kernel, allowed values
1055 are in the range of 100..1000.
1057 cgroup_nodelete=bool Normally fio will delete the cgroups it has created after
1058 the job completion. To override this behavior and to leave
1059 cgroups around after the job completion, set cgroup_nodelete=1.
1060 This can be useful if one wants to inspect various cgroup
1061 files after job completion. Default: false
1063 uid=int Instead of running as the invoking user, set the user ID to
1064 this value before the thread/process does any work.
1066 gid=int Set group ID, see uid.
1068 6.0 Interpreting the output
1069 ---------------------------
1071 fio spits out a lot of output. While running, fio will display the
1072 status of the jobs created. An example of that would be:
1074 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1076 The characters inside the square brackets denote the current status of
1077 each thread. The possible values (in typical life cycle order) are:
1081 P Thread setup, but not started.
1083 I Thread initialized, waiting.
1084 p Thread running pre-reading file(s).
1085 R Running, doing sequential reads.
1086 r Running, doing random reads.
1087 W Running, doing sequential writes.
1088 w Running, doing random writes.
1089 M Running, doing mixed sequential reads/writes.
1090 m Running, doing mixed random reads/writes.
1091 F Running, currently waiting for fsync()
1092 V Running, doing verification of written data.
1093 E Thread exited, not reaped by main thread yet.
1096 The other values are fairly self explanatory - number of threads
1097 currently running and doing io, rate of io since last check (read speed
1098 listed first, then write speed), and the estimated completion percentage
1099 and time for the running group. It's impossible to estimate runtime of
1100 the following groups (if any).
1102 When fio is done (or interrupted by ctrl-c), it will show the data for
1103 each thread, group of threads, and disks in that order. For each data
1104 direction, the output looks like:
1106 Client1 (g=0): err= 0:
1107 write: io= 32MB, bw= 666KB/s, runt= 50320msec
1108 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
1109 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
1110 bw (KB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
1111 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
1112 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
1113 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1114 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1115 issued r/w: total=0/32768, short=0/0
1116 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
1117 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
1119 The client number is printed, along with the group id and error of that
1120 thread. Below is the io statistics, here for writes. In the order listed,
1123 io= Number of megabytes io performed
1124 bw= Average bandwidth rate
1125 runt= The runtime of that thread
1126 slat= Submission latency (avg being the average, stdev being the
1127 standard deviation). This is the time it took to submit
1128 the io. For sync io, the slat is really the completion
1129 latency, since queue/complete is one operation there. This
1130 value can be in milliseconds or microseconds, fio will choose
1131 the most appropriate base and print that. In the example
1132 above, milliseconds is the best scale.
1133 clat= Completion latency. Same names as slat, this denotes the
1134 time from submission to completion of the io pieces. For
1135 sync io, clat will usually be equal (or very close) to 0,
1136 as the time from submit to complete is basically just
1137 CPU time (io has already been done, see slat explanation).
1138 bw= Bandwidth. Same names as the xlat stats, but also includes
1139 an approximate percentage of total aggregate bandwidth
1140 this thread received in this group. This last value is
1141 only really useful if the threads in this group are on the
1142 same disk, since they are then competing for disk access.
1143 cpu= CPU usage. User and system time, along with the number
1144 of context switches this thread went through, usage of
1145 system and user time, and finally the number of major
1146 and minor page faults.
1147 IO depths= The distribution of io depths over the job life time. The
1148 numbers are divided into powers of 2, so for example the
1149 16= entries includes depths up to that value but higher
1150 than the previous entry. In other words, it covers the
1151 range from 16 to 31.
1152 IO submit= How many pieces of IO were submitting in a single submit
1153 call. Each entry denotes that amount and below, until
1154 the previous entry - eg, 8=100% mean that we submitted
1155 anywhere in between 5-8 ios per submit call.
1156 IO complete= Like the above submit number, but for completions instead.
1157 IO issued= The number of read/write requests issued, and how many
1159 IO latencies= The distribution of IO completion latencies. This is the
1160 time from when IO leaves fio and when it gets completed.
1161 The numbers follow the same pattern as the IO depths,
1162 meaning that 2=1.6% means that 1.6% of the IO completed
1163 within 2 msecs, 20=12.8% means that 12.8% of the IO
1164 took more than 10 msecs, but less than (or equal to) 20 msecs.
1166 After each client has been listed, the group statistics are printed. They
1167 will look like this:
1169 Run status group 0 (all jobs):
1170 READ: io=64MB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
1171 WRITE: io=64MB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
1173 For each data direction, it prints:
1175 io= Number of megabytes io performed.
1176 aggrb= Aggregate bandwidth of threads in this group.
1177 minb= The minimum average bandwidth a thread saw.
1178 maxb= The maximum average bandwidth a thread saw.
1179 mint= The smallest runtime of the threads in that group.
1180 maxt= The longest runtime of the threads in that group.
1182 And finally, the disk statistics are printed. They will look like this:
1184 Disk stats (read/write):
1185 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
1187 Each value is printed for both reads and writes, with reads first. The
1190 ios= Number of ios performed by all groups.
1191 merge= Number of merges io the io scheduler.
1192 ticks= Number of ticks we kept the disk busy.
1193 io_queue= Total time spent in the disk queue.
1194 util= The disk utilization. A value of 100% means we kept the disk
1195 busy constantly, 50% would be a disk idling half of the time.
1201 For scripted usage where you typically want to generate tables or graphs
1202 of the results, fio can output the results in a semicolon separated format.
1203 The format is one long line of values, such as:
1205 client1;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%
1206 ;0.0%;0.0%;0.0%;0.0%;0.0%
1208 To enable terse output, use the --minimal command line option.
1210 Split up, the format is as follows:
1212 jobname, groupid, error
1214 KB IO, bandwidth (KB/sec), runtime (msec)
1215 Submission latency: min, max, mean, deviation
1216 Completion latency: min, max, mean, deviation
1217 Bw: min, max, aggregate percentage of total, mean, deviation
1219 KB IO, bandwidth (KB/sec), runtime (msec)
1220 Submission latency: min, max, mean, deviation
1221 Completion latency: min, max, mean, deviation
1222 Bw: min, max, aggregate percentage of total, mean, deviation
1223 CPU usage: user, system, context switches, major faults, minor faults
1224 IO depths: <=1, 2, 4, 8, 16, 32, >=64
1225 IO latencies: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, >=2000