8 5. Detailed list of parameters
12 9. CPU idleness profiling
14 1.0 Overview and history
15 ------------------------
16 fio was originally written to save me the hassle of writing special test
17 case programs when I wanted to test a specific workload, either for
18 performance reasons or to find/reproduce a bug. The process of writing
19 such a test app can be tiresome, especially if you have to do it often.
20 Hence I needed a tool that would be able to simulate a given io workload
21 without resorting to writing a tailored test case again and again.
23 A test work load is difficult to define, though. There can be any number
24 of processes or threads involved, and they can each be using their own
25 way of generating io. You could have someone dirtying large amounts of
26 memory in an memory mapped file, or maybe several threads issuing
27 reads using asynchronous io. fio needed to be flexible enough to
28 simulate both of these cases, and many more.
32 The first step in getting fio to simulate a desired io workload, is
33 writing a job file describing that specific setup. A job file may contain
34 any number of threads and/or files - the typical contents of the job file
35 is a global section defining shared parameters, and one or more job
36 sections describing the jobs involved. When run, fio parses this file
37 and sets everything up as described. If we break down a job from top to
38 bottom, it contains the following basic parameters:
40 IO type Defines the io pattern issued to the file(s).
41 We may only be reading sequentially from this
42 file(s), or we may be writing randomly. Or even
43 mixing reads and writes, sequentially or randomly.
45 Block size In how large chunks are we issuing io? This may be
46 a single value, or it may describe a range of
49 IO size How much data are we going to be reading/writing.
51 IO engine How do we issue io? We could be memory mapping the
52 file, we could be using regular read/write, we
53 could be using splice, async io, syslet, or even
56 IO depth If the io engine is async, how large a queuing
57 depth do we want to maintain?
59 IO type Should we be doing buffered io, or direct/raw io?
61 Num files How many files are we spreading the workload over.
63 Num threads How many threads or processes should we spread
66 The above are the basic parameters defined for a workload, in addition
67 there's a multitude of parameters that modify other aspects of how this
73 See the README file for command line parameters, there are only a few
76 Running fio is normally the easiest part - you just give it the job file
77 (or job files) as parameters:
81 and it will start doing what the job_file tells it to do. You can give
82 more than one job file on the command line, fio will serialize the running
83 of those files. Internally that is the same as using the 'stonewall'
84 parameter described the the parameter section.
86 If the job file contains only one job, you may as well just give the
87 parameters on the command line. The command line parameters are identical
88 to the job parameters, with a few extra that control global parameters
89 (see README). For example, for the job file parameter iodepth=2, the
90 mirror command line option would be --iodepth 2 or --iodepth=2. You can
91 also use the command line for giving more than one job entry. For each
92 --name option that fio sees, it will start a new job with that name.
93 Command line entries following a --name entry will apply to that job,
94 until there are no more entries or a new --name entry is seen. This is
95 similar to the job file options, where each option applies to the current
96 job until a new [] job entry is seen.
98 fio does not need to run as root, except if the files or devices specified
99 in the job section requires that. Some other options may also be restricted,
100 such as memory locking, io scheduler switching, and decreasing the nice value.
105 As previously described, fio accepts one or more job files describing
106 what it is supposed to do. The job file format is the classic ini file,
107 where the names enclosed in [] brackets define the job name. You are free
108 to use any ascii name you want, except 'global' which has special meaning.
109 A global section sets defaults for the jobs described in that file. A job
110 may override a global section parameter, and a job file may even have
111 several global sections if so desired. A job is only affected by a global
112 section residing above it. If the first character in a line is a ';' or a
113 '#', the entire line is discarded as a comment.
115 So let's look at a really simple job file that defines two processes, each
116 randomly reading from a 128MB file.
118 ; -- start job file --
129 As you can see, the job file sections themselves are empty as all the
130 described parameters are shared. As no filename= option is given, fio
131 makes up a filename for each of the jobs as it sees fit. On the command
132 line, this job would look as follows:
134 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
137 Let's look at an example that has a number of processes writing randomly
140 ; -- start job file --
152 Here we have no global section, as we only have one job defined anyway.
153 We want to use async io here, with a depth of 4 for each file. We also
154 increased the buffer size used to 32KB and define numjobs to 4 to
155 fork 4 identical jobs. The result is 4 processes each randomly writing
156 to their own 64MB file. Instead of using the above job file, you could
157 have given the parameters on the command line. For this case, you would
160 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
162 4.1 Environment variables
163 -------------------------
165 fio also supports environment variable expansion in job files. Any
166 substring of the form "${VARNAME}" as part of an option value (in other
167 words, on the right of the `='), will be expanded to the value of the
168 environment variable called VARNAME. If no such environment variable
169 is defined, or VARNAME is the empty string, the empty string will be
172 As an example, let's look at a sample fio invocation and job file:
174 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
176 ; -- start job file --
183 This will expand to the following equivalent job file at runtime:
185 ; -- start job file --
192 fio ships with a few example job files, you can also look there for
195 4.2 Reserved keywords
196 ---------------------
198 Additionally, fio has a set of reserved keywords that will be replaced
199 internally with the appropriate value. Those keywords are:
201 $pagesize The architecture page size of the running system
202 $mb_memory Megabytes of total memory in the system
203 $ncpus Number of online available CPUs
205 These can be used on the command line or in the job file, and will be
206 automatically substituted with the current system values when the job
207 is run. Simple math is also supported on these keywords, so you can
208 perform actions like:
212 and get that properly expanded to 8 times the size of memory in the
216 5.0 Detailed list of parameters
217 -------------------------------
219 This section describes in details each parameter associated with a job.
220 Some parameters take an option of a given type, such as an integer or
221 a string. The following types are used:
223 str String. This is a sequence of alpha characters.
224 time Integer with possible time suffix. In seconds unless otherwise
225 specified, use eg 10m for 10 minutes. Accepts s/m/h for seconds,
227 int SI integer. A whole number value, which may contain a suffix
228 describing the base of the number. Accepted suffixes are k/m/g/t/p,
229 meaning kilo, mega, giga, tera, and peta. The suffix is not case
230 sensitive, and you may also include trailing 'b' (eg 'kb' is the same
231 as 'k'). So if you want to specify 4096, you could either write
232 out '4096' or just give 4k. The suffixes signify base 2 values, so
233 1024 is 1k and 1024k is 1m and so on, unless the suffix is explicitly
234 set to a base 10 value using 'kib', 'mib', 'gib', etc. If that is the
235 case, then 1000 is used as the multiplier. This can be handy for
236 disks, since manufacturers generally use base 10 values when listing
237 the capacity of a drive. If the option accepts an upper and lower
238 range, use a colon ':' or minus '-' to separate such values. May also
239 include a prefix to indicate numbers base. If 0x is used, the number
240 is assumed to be hexadecimal. See irange.
241 bool Boolean. Usually parsed as an integer, however only defined for
242 true and false (1 and 0).
243 irange Integer range with suffix. Allows value range to be given, such
244 as 1024-4096. A colon may also be used as the separator, eg
245 1k:4k. If the option allows two sets of ranges, they can be
246 specified with a ',' or '/' delimiter: 1k-4k/8k-32k. Also see
248 float_list A list of floating numbers, separated by a ':' character.
250 With the above in mind, here follows the complete list of fio job
253 name=str ASCII name of the job. This may be used to override the
254 name printed by fio for this job. Otherwise the job
255 name is used. On the command line this parameter has the
256 special purpose of also signaling the start of a new
259 description=str Text description of the job. Doesn't do anything except
260 dump this text description when this job is run. It's
263 directory=str Prefix filenames with this directory. Used to place files
264 in a different location than "./".
266 filename=str Fio normally makes up a filename based on the job name,
267 thread number, and file number. If you want to share
268 files between threads in a job or several jobs, specify
269 a filename for each of them to override the default. If
270 the ioengine used is 'net', the filename is the host, port,
271 and protocol to use in the format of =host,port,protocol.
272 See ioengine=net for more. If the ioengine is file based, you
273 can specify a number of files by separating the names with a
274 ':' colon. So if you wanted a job to open /dev/sda and /dev/sdb
275 as the two working files, you would use
276 filename=/dev/sda:/dev/sdb. On Windows, disk devices are
277 accessed as \\.\PhysicalDrive0 for the first device,
278 \\.\PhysicalDrive1 for the second etc. Note: Windows and
279 FreeBSD prevent write access to areas of the disk containing
280 in-use data (e.g. filesystems).
281 If the wanted filename does need to include a colon, then
282 escape that with a '\' character. For instance, if the filename
283 is "/dev/dsk/foo@3,0:c", then you would use
284 filename="/dev/dsk/foo@3,0\:c". '-' is a reserved name, meaning
285 stdin or stdout. Which of the two depends on the read/write
289 If sharing multiple files between jobs, it is usually necessary
290 to have fio generate the exact names that you want. By default,
291 fio will name a file based on the default file format
292 specification of jobname.jobnumber.filenumber. With this
293 option, that can be customized. Fio will recognize and replace
294 the following keywords in this string:
297 The name of the worker thread or process.
300 The incremental number of the worker thread or
304 The incremental number of the file for that worker
307 To have dependent jobs share a set of files, this option can
308 be set to have fio generate filenames that are shared between
309 the two. For instance, if testfiles.$filenum is specified,
310 file number 4 for any job will be named testfiles.4. The
311 default of $jobname.$jobnum.$filenum will be used if
312 no other format specifier is given.
314 opendir=str Tell fio to recursively add any file it can find in this
315 directory and down the file system tree.
317 lockfile=str Fio defaults to not locking any files before it does
318 IO to them. If a file or file descriptor is shared, fio
319 can serialize IO to that file to make the end result
320 consistent. This is usual for emulating real workloads that
321 share files. The lock modes are:
323 none No locking. The default.
324 exclusive Only one thread/process may do IO,
325 excluding all others.
326 readwrite Read-write locking on the file. Many
327 readers may access the file at the
328 same time, but writes get exclusive
332 rw=str Type of io pattern. Accepted values are:
334 read Sequential reads
335 write Sequential writes
336 randwrite Random writes
337 randread Random reads
338 rw,readwrite Sequential mixed reads and writes
339 randrw Random mixed reads and writes
341 For the mixed io types, the default is to split them 50/50.
342 For certain types of io the result may still be skewed a bit,
343 since the speed may be different. It is possible to specify
344 a number of IO's to do before getting a new offset, this is
345 one by appending a ':<nr>' to the end of the string given.
346 For a random read, it would look like 'rw=randread:8' for
347 passing in an offset modifier with a value of 8. If the
348 suffix is used with a sequential IO pattern, then the value
349 specified will be added to the generated offset for each IO.
350 For instance, using rw=write:4k will skip 4k for every
351 write. It turns sequential IO into sequential IO with holes.
352 See the 'rw_sequencer' option.
354 rw_sequencer=str If an offset modifier is given by appending a number to
355 the rw=<str> line, then this option controls how that
356 number modifies the IO offset being generated. Accepted
359 sequential Generate sequential offset
360 identical Generate the same offset
362 'sequential' is only useful for random IO, where fio would
363 normally generate a new random offset for every IO. If you
364 append eg 8 to randread, you would get a new random offset for
365 every 8 IO's. The result would be a seek for only every 8
366 IO's, instead of for every IO. Use rw=randread:8 to specify
367 that. As sequential IO is already sequential, setting
368 'sequential' for that would not result in any differences.
369 'identical' behaves in a similar fashion, except it sends
370 the same offset 8 number of times before generating a new
373 kb_base=int The base unit for a kilobyte. The defacto base is 2^10, 1024.
374 Storage manufacturers like to use 10^3 or 1000 as a base
375 ten unit instead, for obvious reasons. Allow values are
376 1024 or 1000, with 1024 being the default.
378 unified_rw_reporting=bool Fio normally reports statistics on a per
379 data direction basis, meaning that read, write, and trim are
380 accounted and reported separately. If this option is set,
381 the fio will sum the results and report them as "mixed"
384 randrepeat=bool For random IO workloads, seed the generator in a predictable
385 way so that results are repeatable across repetitions.
387 use_os_rand=bool Fio can either use the random generator supplied by the OS
388 to generator random offsets, or it can use it's own internal
389 generator (based on Tausworthe). Default is to use the
390 internal generator, which is often of better quality and
393 fallocate=str Whether pre-allocation is performed when laying down files.
396 none Do not pre-allocate space
397 posix Pre-allocate via posix_fallocate()
398 keep Pre-allocate via fallocate() with
399 FALLOC_FL_KEEP_SIZE set
400 0 Backward-compatible alias for 'none'
401 1 Backward-compatible alias for 'posix'
403 May not be available on all supported platforms. 'keep' is only
404 available on Linux.If using ZFS on Solaris this must be set to
405 'none' because ZFS doesn't support it. Default: 'posix'.
407 fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
408 on what IO patterns it is likely to issue. Sometimes you
409 want to test specific IO patterns without telling the
410 kernel about it, in which case you can disable this option.
411 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
412 IO and POSIX_FADV_RANDOM for random IO.
414 size=int The total size of file io for this job. Fio will run until
415 this many bytes has been transferred, unless runtime is
416 limited by other options (such as 'runtime', for instance).
417 Unless specific nrfiles and filesize options are given,
418 fio will divide this size between the available files
419 specified by the job. If not set, fio will use the full
420 size of the given files or devices. If the the files
421 do not exist, size must be given. It is also possible to
422 give size as a percentage between 1 and 100. If size=20%
423 is given, fio will use 20% of the full size of the given
426 filesize=int Individual file sizes. May be a range, in which case fio
427 will select sizes for files at random within the given range
428 and limited to 'size' in total (if that is given). If not
429 given, each created file is the same size.
432 fill_fs=bool Sets size to something really large and waits for ENOSPC (no
433 space left on device) as the terminating condition. Only makes
434 sense with sequential write. For a read workload, the mount
435 point will be filled first then IO started on the result. This
436 option doesn't make sense if operating on a raw device node,
437 since the size of that is already known by the file system.
438 Additionally, writing beyond end-of-device will not return
442 bs=int The block size used for the io units. Defaults to 4k. Values
443 can be given for both read and writes. If a single int is
444 given, it will apply to both. If a second int is specified
445 after a comma, it will apply to writes only. In other words,
446 the format is either bs=read_and_write or bs=read,write,trim.
447 bs=4k,8k will thus use 4k blocks for reads, 8k blocks for
448 writes, and 8k for trims. You can terminate the list with
449 a trailing comma. bs=4k,8k, would use the default value for
450 trims.. If you only wish to set the write size, you
451 can do so by passing an empty read size - bs=,8k will set
452 8k for writes and leave the read default value.
455 ba=int At what boundary to align random IO offsets. Defaults to
456 the same as 'blocksize' the minimum blocksize given.
457 Minimum alignment is typically 512b for using direct IO,
458 though it usually depends on the hardware block size. This
459 option is mutually exclusive with using a random map for
460 files, so it will turn off that option.
462 blocksize_range=irange
463 bsrange=irange Instead of giving a single block size, specify a range
464 and fio will mix the issued io block sizes. The issued
465 io unit will always be a multiple of the minimum value
466 given (also see bs_unaligned). Applies to both reads and
467 writes, however a second range can be given after a comma.
470 bssplit=str Sometimes you want even finer grained control of the
471 block sizes issued, not just an even split between them.
472 This option allows you to weight various block sizes,
473 so that you are able to define a specific amount of
474 block sizes issued. The format for this option is:
476 bssplit=blocksize/percentage:blocksize/percentage
478 for as many block sizes as needed. So if you want to define
479 a workload that has 50% 64k blocks, 10% 4k blocks, and
480 40% 32k blocks, you would write:
482 bssplit=4k/10:64k/50:32k/40
484 Ordering does not matter. If the percentage is left blank,
485 fio will fill in the remaining values evenly. So a bssplit
486 option like this one:
488 bssplit=4k/50:1k/:32k/
490 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
491 always add up to 100, if bssplit is given a range that adds
492 up to more, it will error out.
494 bssplit also supports giving separate splits to reads and
495 writes. The format is identical to what bs= accepts. You
496 have to separate the read and write parts with a comma. So
497 if you want a workload that has 50% 2k reads and 50% 4k reads,
498 while having 90% 4k writes and 10% 8k writes, you would
501 bssplit=2k/50:4k/50,4k/90,8k/10
504 bs_unaligned If this option is given, any byte size value within bsrange
505 may be used as a block range. This typically wont work with
506 direct IO, as that normally requires sector alignment.
508 bs_is_seq_rand If this option is set, fio will use the normal read,write
509 blocksize settings as sequential,random instead. Any random
510 read or write will use the WRITE blocksize settings, and any
511 sequential read or write will use the READ blocksize setting.
513 zero_buffers If this option is given, fio will init the IO buffers to
514 all zeroes. The default is to fill them with random data.
516 refill_buffers If this option is given, fio will refill the IO buffers
517 on every submit. The default is to only fill it at init
518 time and reuse that data. Only makes sense if zero_buffers
519 isn't specified, naturally. If data verification is enabled,
520 refill_buffers is also automatically enabled.
522 scramble_buffers=bool If refill_buffers is too costly and the target is
523 using data deduplication, then setting this option will
524 slightly modify the IO buffer contents to defeat normal
525 de-dupe attempts. This is not enough to defeat more clever
526 block compression attempts, but it will stop naive dedupe of
527 blocks. Default: true.
529 buffer_compress_percentage=int If this is set, then fio will attempt to
530 provide IO buffer content (on WRITEs) that compress to
531 the specified level. Fio does this by providing a mix of
532 random data and zeroes. Note that this is per block size
533 unit, for file/disk wide compression level that matches
534 this setting, you'll also want to set refill_buffers.
536 buffer_compress_chunk=int See buffer_compress_percentage. This
537 setting allows fio to manage how big the ranges of random
538 data and zeroed data is. Without this set, fio will
539 provide buffer_compress_percentage of blocksize random
540 data, followed by the remaining zeroed. With this set
541 to some chunk size smaller than the block size, fio can
542 alternate random and zeroed data throughout the IO
545 nrfiles=int Number of files to use for this job. Defaults to 1.
547 openfiles=int Number of files to keep open at the same time. Defaults to
548 the same as nrfiles, can be set smaller to limit the number
551 file_service_type=str Defines how fio decides which file from a job to
552 service next. The following types are defined:
554 random Just choose a file at random.
556 roundrobin Round robin over open files. This
559 sequential Finish one file before moving on to
560 the next. Multiple files can still be
561 open depending on 'openfiles'.
563 The string can have a number appended, indicating how
564 often to switch to a new file. So if option random:4 is
565 given, fio will switch to a new random file after 4 ios
568 ioengine=str Defines how the job issues io to the file. The following
571 sync Basic read(2) or write(2) io. lseek(2) is
572 used to position the io location.
574 psync Basic pread(2) or pwrite(2) io.
576 vsync Basic readv(2) or writev(2) IO.
578 psyncv Basic preadv(2) or pwritev(2) IO.
580 libaio Linux native asynchronous io. Note that Linux
581 may only support queued behaviour with
582 non-buffered IO (set direct=1 or buffered=0).
583 This engine defines engine specific options.
585 posixaio glibc posix asynchronous io.
587 solarisaio Solaris native asynchronous io.
589 windowsaio Windows native asynchronous io.
591 mmap File is memory mapped and data copied
592 to/from using memcpy(3).
594 splice splice(2) is used to transfer the data and
595 vmsplice(2) to transfer data from user
598 syslet-rw Use the syslet system calls to make
599 regular read/write async.
601 sg SCSI generic sg v3 io. May either be
602 synchronous using the SG_IO ioctl, or if
603 the target is an sg character device
604 we use read(2) and write(2) for asynchronous
607 null Doesn't transfer any data, just pretends
608 to. This is mainly used to exercise fio
609 itself and for debugging/testing purposes.
611 net Transfer over the network to given host:port.
612 Depending on the protocol used, the hostname,
613 port, listen and filename options are used to
614 specify what sort of connection to make, while
615 the protocol option determines which protocol
617 This engine defines engine specific options.
619 netsplice Like net, but uses splice/vmsplice to
620 map data and send/receive.
621 This engine defines engine specific options.
623 cpuio Doesn't transfer any data, but burns CPU
624 cycles according to the cpuload= and
625 cpucycle= options. Setting cpuload=85
626 will cause that job to do nothing but burn
627 85% of the CPU. In case of SMP machines,
628 use numjobs=<no_of_cpu> to get desired CPU
629 usage, as the cpuload only loads a single
630 CPU at the desired rate.
632 guasi The GUASI IO engine is the Generic Userspace
633 Asyncronous Syscall Interface approach
636 http://www.xmailserver.org/guasi-lib.html
638 for more info on GUASI.
640 rdma The RDMA I/O engine supports both RDMA
641 memory semantics (RDMA_WRITE/RDMA_READ) and
642 channel semantics (Send/Recv) for the
643 InfiniBand, RoCE and iWARP protocols.
645 falloc IO engine that does regular fallocate to
646 simulate data transfer as fio ioengine.
647 DDIR_READ does fallocate(,mode = keep_size,)
648 DDIR_WRITE does fallocate(,mode = 0)
649 DDIR_TRIM does fallocate(,mode = punch_hole)
651 e4defrag IO engine that does regular EXT4_IOC_MOVE_EXT
652 ioctls to simulate defragment activity in
653 request to DDIR_WRITE event
655 external Prefix to specify loading an external
656 IO engine object file. Append the engine
657 filename, eg ioengine=external:/tmp/foo.o
658 to load ioengine foo.o in /tmp.
660 iodepth=int This defines how many io units to keep in flight against
661 the file. The default is 1 for each file defined in this
662 job, can be overridden with a larger value for higher
663 concurrency. Note that increasing iodepth beyond 1 will not
664 affect synchronous ioengines (except for small degress when
665 verify_async is in use). Even async engines may impose OS
666 restrictions causing the desired depth not to be achieved.
667 This may happen on Linux when using libaio and not setting
668 direct=1, since buffered IO is not async on that OS. Keep an
669 eye on the IO depth distribution in the fio output to verify
670 that the achieved depth is as expected. Default: 1.
672 iodepth_batch_submit=int
673 iodepth_batch=int This defines how many pieces of IO to submit at once.
674 It defaults to 1 which means that we submit each IO
675 as soon as it is available, but can be raised to submit
676 bigger batches of IO at the time.
678 iodepth_batch_complete=int This defines how many pieces of IO to retrieve
679 at once. It defaults to 1 which means that we'll ask
680 for a minimum of 1 IO in the retrieval process from
681 the kernel. The IO retrieval will go on until we
682 hit the limit set by iodepth_low. If this variable is
683 set to 0, then fio will always check for completed
684 events before queuing more IO. This helps reduce
685 IO latency, at the cost of more retrieval system calls.
687 iodepth_low=int The low water mark indicating when to start filling
688 the queue again. Defaults to the same as iodepth, meaning
689 that fio will attempt to keep the queue full at all times.
690 If iodepth is set to eg 16 and iodepth_low is set to 4, then
691 after fio has filled the queue of 16 requests, it will let
692 the depth drain down to 4 before starting to fill it again.
694 direct=bool If value is true, use non-buffered io. This is usually
695 O_DIRECT. Note that ZFS on Solaris doesn't support direct io.
696 On Windows the synchronous ioengines don't support direct io.
698 buffered=bool If value is true, use buffered io. This is the opposite
699 of the 'direct' option. Defaults to true.
701 offset=int Start io at the given offset in the file. The data before
702 the given offset will not be touched. This effectively
703 caps the file size at real_size - offset.
705 offset_increment=int If this is provided, then the real offset becomes
706 the offset + offset_increment * thread_number, where the
707 thread number is a counter that starts at 0 and is incremented
708 for each job. This option is useful if there are several jobs
709 which are intended to operate on a file in parallel in disjoint
710 segments, with even spacing between the starting points.
712 number_ios=int Fio will normally perform IOs until it has exhausted the size
713 of the region set by size=, or if it exhaust the allocated
714 time (or hits an error condition). With this setting, the
715 range/size can be set independently of the number of IOs to
716 perform. When fio reaches this number, it will exit normally
719 fsync=int If writing to a file, issue a sync of the dirty data
720 for every number of blocks given. For example, if you give
721 32 as a parameter, fio will sync the file for every 32
722 writes issued. If fio is using non-buffered io, we may
723 not sync the file. The exception is the sg io engine, which
724 synchronizes the disk cache anyway.
726 fdatasync=int Like fsync= but uses fdatasync() to only sync data and not
728 In FreeBSD and Windows there is no fdatasync(), this falls back to
731 sync_file_range=str:val Use sync_file_range() for every 'val' number of
732 write operations. Fio will track range of writes that
733 have happened since the last sync_file_range() call. 'str'
734 can currently be one or more of:
736 wait_before SYNC_FILE_RANGE_WAIT_BEFORE
737 write SYNC_FILE_RANGE_WRITE
738 wait_after SYNC_FILE_RANGE_WAIT_AFTER
740 So if you do sync_file_range=wait_before,write:8, fio would
741 use SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE for
742 every 8 writes. Also see the sync_file_range(2) man page.
743 This option is Linux specific.
745 overwrite=bool If true, writes to a file will always overwrite existing
746 data. If the file doesn't already exist, it will be
747 created before the write phase begins. If the file exists
748 and is large enough for the specified write phase, nothing
751 end_fsync=bool If true, fsync file contents when a write stage has completed.
753 fsync_on_close=bool If true, fio will fsync() a dirty file on close.
754 This differs from end_fsync in that it will happen on every
755 file close, not just at the end of the job.
757 rwmixread=int How large a percentage of the mix should be reads.
759 rwmixwrite=int How large a percentage of the mix should be writes. If both
760 rwmixread and rwmixwrite is given and the values do not add
761 up to 100%, the latter of the two will be used to override
762 the first. This may interfere with a given rate setting,
763 if fio is asked to limit reads or writes to a certain rate.
764 If that is the case, then the distribution may be skewed.
766 random_distribution=str:float By default, fio will use a completely uniform
767 random distribution when asked to perform random IO. Sometimes
768 it is useful to skew the distribution in specific ways,
769 ensuring that some parts of the data is more hot than others.
770 fio includes the following distribution models:
772 random Uniform random distribution
773 zipf Zipf distribution
774 pareto Pareto distribution
776 When using a zipf or pareto distribution, an input value
777 is also needed to define the access pattern. For zipf, this
778 is the zipf theta. For pareto, it's the pareto power. Fio
779 includes a test program, genzipf, that can be used visualize
780 what the given input values will yield in terms of hit rates.
781 If you wanted to use zipf with a theta of 1.2, you would use
782 random_distribution=zipf:1.2 as the option. If a non-uniform
783 model is used, fio will disable use of the random map.
785 percentage_random=int For a random workload, set how big a percentage should
786 be random. This defaults to 100%, in which case the workload
787 is fully random. It can be set from anywhere from 0 to 100.
788 Setting it to 0 would make the workload fully sequential. Any
789 setting in between will result in a random mix of sequential
790 and random IO, at the given percentages. It is possible to
791 set different values for reads, writes, and trim. To do so,
792 simply use a comma separated list. See blocksize.
794 norandommap Normally fio will cover every block of the file when doing
795 random IO. If this option is given, fio will just get a
796 new random offset without looking at past io history. This
797 means that some blocks may not be read or written, and that
798 some blocks may be read/written more than once. This option
799 is mutually exclusive with verify= if and only if multiple
800 blocksizes (via bsrange=) are used, since fio only tracks
801 complete rewrites of blocks.
803 softrandommap=bool See norandommap. If fio runs with the random block map
804 enabled and it fails to allocate the map, if this option is
805 set it will continue without a random block map. As coverage
806 will not be as complete as with random maps, this option is
809 random_generator=str Fio supports the following engines for generating
810 IO offsets for random IO:
812 tausworthe Strong 2^88 cycle random number generator
813 lfsr Linear feedback shift register generator
815 Tausworthe is a strong random number generator, but it
816 requires tracking on the side if we want to ensure that
817 blocks are only read or written once. LFSR guarantees
818 that we never generate the same offset twice, and it's
819 also less computationally expensive. It's not a true
820 random generator, however, though for IO purposes it's
821 typically good enough. LFSR only works with single
822 block sizes, not with workloads that use multiple block
823 sizes. If used with such a workload, fio may read or write
824 some blocks multiple times.
826 nice=int Run the job with the given nice value. See man nice(2).
828 prio=int Set the io priority value of this job. Linux limits us to
829 a positive value between 0 and 7, with 0 being the highest.
832 prioclass=int Set the io priority class. See man ionice(1).
834 thinktime=int Stall the job x microseconds after an io has completed before
835 issuing the next. May be used to simulate processing being
836 done by an application. See thinktime_blocks and
840 Only valid if thinktime is set - pretend to spend CPU time
841 doing something with the data received, before falling back
842 to sleeping for the rest of the period specified by
846 Only valid if thinktime is set - control how many blocks
847 to issue, before waiting 'thinktime' usecs. If not set,
848 defaults to 1 which will make fio wait 'thinktime' usecs
849 after every block. This effectively makes any queue depth
850 setting redundant, since no more than 1 IO will be queued
851 before we have to complete it and do our thinktime. In
852 other words, this setting effectively caps the queue depth
853 if the latter is larger.
855 rate=int Cap the bandwidth used by this job. The number is in bytes/sec,
856 the normal suffix rules apply. You can use rate=500k to limit
857 reads and writes to 500k each, or you can specify read and
858 writes separately. Using rate=1m,500k would limit reads to
859 1MB/sec and writes to 500KB/sec. Capping only reads or
860 writes can be done with rate=,500k or rate=500k,. The former
861 will only limit writes (to 500KB/sec), the latter will only
864 ratemin=int Tell fio to do whatever it can to maintain at least this
865 bandwidth. Failing to meet this requirement, will cause
866 the job to exit. The same format as rate is used for
867 read vs write separation.
869 rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
870 as rate, just specified independently of bandwidth. If the
871 job is given a block size range instead of a fixed value,
872 the smallest block size is used as the metric. The same format
873 as rate is used for read vs write separation.
875 rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
876 the job to exit. The same format as rate is used for read vs
879 max_latency=int If set, fio will exit the job if it exceeds this maximum
880 latency. It will exit with an ETIME error.
882 ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
885 cpumask=int Set the CPU affinity of this job. The parameter given is a
886 bitmask of allowed CPU's the job may run on. So if you want
887 the allowed CPUs to be 1 and 5, you would pass the decimal
888 value of (1 << 1 | 1 << 5), or 34. See man
889 sched_setaffinity(2). This may not work on all supported
890 operating systems or kernel versions. This option doesn't
891 work well for a higher CPU count than what you can store in
892 an integer mask, so it can only control cpus 1-32. For
893 boxes with larger CPU counts, use cpus_allowed.
895 cpus_allowed=str Controls the same options as cpumask, but it allows a text
896 setting of the permitted CPUs instead. So to use CPUs 1 and
897 5, you would specify cpus_allowed=1,5. This options also
898 allows a range of CPUs. Say you wanted a binding to CPUs
899 1, 5, and 8-15, you would set cpus_allowed=1,5,8-15.
901 numa_cpu_nodes=str Set this job running on spcified NUMA nodes' CPUs. The
902 arguments allow comma delimited list of cpu numbers,
903 A-B ranges, or 'all'. Note, to enable numa options support,
904 fio must be built on a system with libnuma-dev(el) installed.
906 numa_mem_policy=str Set this job's memory policy and corresponding NUMA
907 nodes. Format of the argements:
909 `mode' is one of the following memory policy:
910 default, prefer, bind, interleave, local
911 For `default' and `local' memory policy, no node is
912 needed to be specified.
913 For `prefer', only one node is allowed.
914 For `bind' and `interleave', it allow comma delimited
915 list of numbers, A-B ranges, or 'all'.
917 startdelay=time Start this job the specified number of seconds after fio
918 has started. Only useful if the job file contains several
919 jobs, and you want to delay starting some jobs to a certain
922 runtime=time Tell fio to terminate processing after the specified number
923 of seconds. It can be quite hard to determine for how long
924 a specified job will run, so this parameter is handy to
925 cap the total runtime to a given time.
927 time_based If set, fio will run for the duration of the runtime
928 specified even if the file(s) are completely read or
929 written. It will simply loop over the same workload
930 as many times as the runtime allows.
932 ramp_time=time If set, fio will run the specified workload for this amount
933 of time before logging any performance numbers. Useful for
934 letting performance settle before logging results, thus
935 minimizing the runtime required for stable results. Note
936 that the ramp_time is considered lead in time for a job,
937 thus it will increase the total runtime if a special timeout
938 or runtime is specified.
940 invalidate=bool Invalidate the buffer/page cache parts for this file prior
941 to starting io. Defaults to true.
943 sync=bool Use sync io for buffered writes. For the majority of the
944 io engines, this means using O_SYNC.
947 mem=str Fio can use various types of memory as the io unit buffer.
948 The allowed values are:
950 malloc Use memory from malloc(3) as the buffers.
952 shm Use shared memory as the buffers. Allocated
955 shmhuge Same as shm, but use huge pages as backing.
957 mmap Use mmap to allocate buffers. May either be
958 anonymous memory, or can be file backed if
959 a filename is given after the option. The
960 format is mem=mmap:/path/to/file.
962 mmaphuge Use a memory mapped huge file as the buffer
963 backing. Append filename after mmaphuge, ala
964 mem=mmaphuge:/hugetlbfs/file
966 The area allocated is a function of the maximum allowed
967 bs size for the job, multiplied by the io depth given. Note
968 that for shmhuge and mmaphuge to work, the system must have
969 free huge pages allocated. This can normally be checked
970 and set by reading/writing /proc/sys/vm/nr_hugepages on a
971 Linux system. Fio assumes a huge page is 4MB in size. So
972 to calculate the number of huge pages you need for a given
973 job file, add up the io depth of all jobs (normally one unless
974 iodepth= is used) and multiply by the maximum bs set. Then
975 divide that number by the huge page size. You can see the
976 size of the huge pages in /proc/meminfo. If no huge pages
977 are allocated by having a non-zero number in nr_hugepages,
978 using mmaphuge or shmhuge will fail. Also see hugepage-size.
980 mmaphuge also needs to have hugetlbfs mounted and the file
981 location should point there. So if it's mounted in /huge,
982 you would use mem=mmaphuge:/huge/somefile.
984 iomem_align=int This indiciates the memory alignment of the IO memory buffers.
985 Note that the given alignment is applied to the first IO unit
986 buffer, if using iodepth the alignment of the following buffers
987 are given by the bs used. In other words, if using a bs that is
988 a multiple of the page sized in the system, all buffers will
989 be aligned to this value. If using a bs that is not page
990 aligned, the alignment of subsequent IO memory buffers is the
991 sum of the iomem_align and bs used.
994 Defines the size of a huge page. Must at least be equal
995 to the system setting, see /proc/meminfo. Defaults to 4MB.
996 Should probably always be a multiple of megabytes, so using
997 hugepage-size=Xm is the preferred way to set this to avoid
998 setting a non-pow-2 bad value.
1000 exitall When one job finishes, terminate the rest. The default is
1001 to wait for each job to finish, sometimes that is not the
1004 bwavgtime=int Average the calculated bandwidth over the given time. Value
1005 is specified in milliseconds.
1007 iopsavgtime=int Average the calculated IOPS over the given time. Value
1008 is specified in milliseconds.
1010 create_serialize=bool If true, serialize the file creating for the jobs.
1011 This may be handy to avoid interleaving of data
1012 files, which may greatly depend on the filesystem
1013 used and even the number of processors in the system.
1015 create_fsync=bool fsync the data file after creation. This is the
1018 create_on_open=bool Don't pre-setup the files for IO, just create open()
1019 when it's time to do IO to that file.
1021 create_only=bool If true, fio will only run the setup phase of the job.
1022 If files need to be laid out or updated on disk, only
1023 that will be done. The actual job contents are not
1026 pre_read=bool If this is given, files will be pre-read into memory before
1027 starting the given IO operation. This will also clear
1028 the 'invalidate' flag, since it is pointless to pre-read
1029 and then drop the cache. This will only work for IO engines
1030 that are seekable, since they allow you to read the same data
1031 multiple times. Thus it will not work on eg network or splice
1034 unlink=bool Unlink the job files when done. Not the default, as repeated
1035 runs of that job would then waste time recreating the file
1036 set again and again.
1038 loops=int Run the specified number of iterations of this job. Used
1039 to repeat the same workload a given number of times. Defaults
1042 do_verify=bool Run the verify phase after a write phase. Only makes sense if
1043 verify is set. Defaults to 1.
1045 verify=str If writing to a file, fio can verify the file contents
1046 after each iteration of the job. The allowed values are:
1048 md5 Use an md5 sum of the data area and store
1049 it in the header of each block.
1051 crc64 Use an experimental crc64 sum of the data
1052 area and store it in the header of each
1055 crc32c Use a crc32c sum of the data area and store
1056 it in the header of each block.
1058 crc32c-intel Use hardware assisted crc32c calcuation
1059 provided on SSE4.2 enabled processors. Falls
1060 back to regular software crc32c, if not
1061 supported by the system.
1063 crc32 Use a crc32 sum of the data area and store
1064 it in the header of each block.
1066 crc16 Use a crc16 sum of the data area and store
1067 it in the header of each block.
1069 crc7 Use a crc7 sum of the data area and store
1070 it in the header of each block.
1072 sha512 Use sha512 as the checksum function.
1074 sha256 Use sha256 as the checksum function.
1076 sha1 Use optimized sha1 as the checksum function.
1078 meta Write extra information about each io
1079 (timestamp, block number etc.). The block
1080 number is verified. See also verify_pattern.
1082 null Only pretend to verify. Useful for testing
1083 internals with ioengine=null, not for much
1086 This option can be used for repeated burn-in tests of a
1087 system to make sure that the written data is also
1088 correctly read back. If the data direction given is
1089 a read or random read, fio will assume that it should
1090 verify a previously written file. If the data direction
1091 includes any form of write, the verify will be of the
1094 verifysort=bool If set, fio will sort written verify blocks when it deems
1095 it faster to read them back in a sorted manner. This is
1096 often the case when overwriting an existing file, since
1097 the blocks are already laid out in the file system. You
1098 can ignore this option unless doing huge amounts of really
1099 fast IO where the red-black tree sorting CPU time becomes
1102 verify_offset=int Swap the verification header with data somewhere else
1103 in the block before writing. Its swapped back before
1106 verify_interval=int Write the verification header at a finer granularity
1107 than the blocksize. It will be written for chunks the
1108 size of header_interval. blocksize should divide this
1111 verify_pattern=str If set, fio will fill the io buffers with this
1112 pattern. Fio defaults to filling with totally random
1113 bytes, but sometimes it's interesting to fill with a known
1114 pattern for io verification purposes. Depending on the
1115 width of the pattern, fio will fill 1/2/3/4 bytes of the
1116 buffer at the time(it can be either a decimal or a hex number).
1117 The verify_pattern if larger than a 32-bit quantity has to
1118 be a hex number that starts with either "0x" or "0X". Use
1121 verify_fatal=bool Normally fio will keep checking the entire contents
1122 before quitting on a block verification failure. If this
1123 option is set, fio will exit the job on the first observed
1126 verify_dump=bool If set, dump the contents of both the original data
1127 block and the data block we read off disk to files. This
1128 allows later analysis to inspect just what kind of data
1129 corruption occurred. Off by default.
1131 verify_async=int Fio will normally verify IO inline from the submitting
1132 thread. This option takes an integer describing how many
1133 async offload threads to create for IO verification instead,
1134 causing fio to offload the duty of verifying IO contents
1135 to one or more separate threads. If using this offload
1136 option, even sync IO engines can benefit from using an
1137 iodepth setting higher than 1, as it allows them to have
1138 IO in flight while verifies are running.
1140 verify_async_cpus=str Tell fio to set the given CPU affinity on the
1141 async IO verification threads. See cpus_allowed for the
1144 verify_backlog=int Fio will normally verify the written contents of a
1145 job that utilizes verify once that job has completed. In
1146 other words, everything is written then everything is read
1147 back and verified. You may want to verify continually
1148 instead for a variety of reasons. Fio stores the meta data
1149 associated with an IO block in memory, so for large
1150 verify workloads, quite a bit of memory would be used up
1151 holding this meta data. If this option is enabled, fio
1152 will write only N blocks before verifying these blocks.
1154 verify_backlog_batch=int Control how many blocks fio will verify
1155 if verify_backlog is set. If not set, will default to
1156 the value of verify_backlog (meaning the entire queue
1157 is read back and verified). If verify_backlog_batch is
1158 less than verify_backlog then not all blocks will be verified,
1159 if verify_backlog_batch is larger than verify_backlog, some
1160 blocks will be verified more than once.
1163 wait_for_previous Wait for preceding jobs in the job file to exit, before
1164 starting this one. Can be used to insert serialization
1165 points in the job file. A stone wall also implies starting
1166 a new reporting group.
1168 new_group Start a new reporting group. See: group_reporting.
1170 numjobs=int Create the specified number of clones of this job. May be
1171 used to setup a larger number of threads/processes doing
1172 the same thing. Each thread is reported separately; to see
1173 statistics for all clones as a whole, use group_reporting in
1174 conjunction with new_group.
1176 group_reporting It may sometimes be interesting to display statistics for
1177 groups of jobs as a whole instead of for each individual job.
1178 This is especially true if 'numjobs' is used; looking at
1179 individual thread/process output quickly becomes unwieldy.
1180 To see the final report per-group instead of per-job, use
1181 'group_reporting'. Jobs in a file will be part of the same
1182 reporting group, unless if separated by a stonewall, or by
1185 thread fio defaults to forking jobs, however if this option is
1186 given, fio will use pthread_create(3) to create threads
1189 zonesize=int Divide a file into zones of the specified size. See zoneskip.
1191 zoneskip=int Skip the specified number of bytes when zonesize data has
1192 been read. The two zone options can be used to only do
1193 io on zones of a file.
1195 write_iolog=str Write the issued io patterns to the specified file. See
1196 read_iolog. Specify a separate file for each job, otherwise
1197 the iologs will be interspersed and the file may be corrupt.
1199 read_iolog=str Open an iolog with the specified file name and replay the
1200 io patterns it contains. This can be used to store a
1201 workload and replay it sometime later. The iolog given
1202 may also be a blktrace binary file, which allows fio
1203 to replay a workload captured by blktrace. See blktrace
1204 for how to capture such logging data. For blktrace replay,
1205 the file needs to be turned into a blkparse binary data
1206 file first (blkparse <device> -o /dev/null -d file_for_fio.bin).
1208 replay_no_stall=int When replaying I/O with read_iolog the default behavior
1209 is to attempt to respect the time stamps within the log and
1210 replay them with the appropriate delay between IOPS. By
1211 setting this variable fio will not respect the timestamps and
1212 attempt to replay them as fast as possible while still
1213 respecting ordering. The result is the same I/O pattern to a
1214 given device, but different timings.
1216 replay_redirect=str While replaying I/O patterns using read_iolog the
1217 default behavior is to replay the IOPS onto the major/minor
1218 device that each IOP was recorded from. This is sometimes
1219 undesirable because on a different machine those major/minor
1220 numbers can map to a different device. Changing hardware on
1221 the same system can also result in a different major/minor
1222 mapping. Replay_redirect causes all IOPS to be replayed onto
1223 the single specified device regardless of the device it was
1224 recorded from. i.e. replay_redirect=/dev/sdc would cause all
1225 IO in the blktrace to be replayed onto /dev/sdc. This means
1226 multiple devices will be replayed onto a single, if the trace
1227 contains multiple devices. If you want multiple devices to be
1228 replayed concurrently to multiple redirected devices you must
1229 blkparse your trace into separate traces and replay them with
1230 independent fio invocations. Unfortuantely this also breaks
1231 the strict time ordering between multiple device accesses.
1233 write_bw_log=str If given, write a bandwidth log of the jobs in this job
1234 file. Can be used to store data of the bandwidth of the
1235 jobs in their lifetime. The included fio_generate_plots
1236 script uses gnuplot to turn these text files into nice
1237 graphs. See write_lat_log for behaviour of given
1238 filename. For this option, the suffix is _bw.log.
1240 write_lat_log=str Same as write_bw_log, except that this option stores io
1241 submission, completion, and total latencies instead. If no
1242 filename is given with this option, the default filename of
1243 "jobname_type.log" is used. Even if the filename is given,
1244 fio will still append the type of log. So if one specifies
1248 The actual log names will be foo_slat.log, foo_clat.log,
1249 and foo_lat.log. This helps fio_generate_plot fine the logs
1252 write_iops_log=str Same as write_bw_log, but writes IOPS. If no filename is
1253 given with this option, the default filename of
1254 "jobname_type.log" is used. Even if the filename is given,
1255 fio will still append the type of log.
1257 log_avg_msec=int By default, fio will log an entry in the iops, latency,
1258 or bw log for every IO that completes. When writing to the
1259 disk log, that can quickly grow to a very large size. Setting
1260 this option makes fio average the each log entry over the
1261 specified period of time, reducing the resolution of the log.
1264 lockmem=int Pin down the specified amount of memory with mlock(2). Can
1265 potentially be used instead of removing memory or booting
1266 with less memory to simulate a smaller amount of memory.
1267 The amount specified is per worker.
1269 exec_prerun=str Before running this job, issue the command specified
1270 through system(3). Output is redirected in a file called
1273 exec_postrun=str After the job completes, issue the command specified
1274 though system(3). Output is redirected in a file called
1275 jobname.postrun.txt.
1277 ioscheduler=str Attempt to switch the device hosting the file to the specified
1278 io scheduler before running.
1280 disk_util=bool Generate disk utilization statistics, if the platform
1281 supports it. Defaults to on.
1283 disable_lat=bool Disable measurements of total latency numbers. Useful
1284 only for cutting back the number of calls to gettimeofday,
1285 as that does impact performance at really high IOPS rates.
1286 Note that to really get rid of a large amount of these
1287 calls, this option must be used with disable_slat and
1290 disable_clat=bool Disable measurements of completion latency numbers. See
1293 disable_slat=bool Disable measurements of submission latency numbers. See
1296 disable_bw=bool Disable measurements of throughput/bandwidth numbers. See
1299 clat_percentiles=bool Enable the reporting of percentiles of
1300 completion latencies.
1302 percentile_list=float_list Overwrite the default list of percentiles
1303 for completion latencies. Each number is a floating
1304 number in the range (0,100], and the maximum length of
1305 the list is 20. Use ':' to separate the numbers, and
1306 list the numbers in ascending order. For example,
1307 --percentile_list=99.5:99.9 will cause fio to report
1308 the values of completion latency below which 99.5% and
1309 99.9% of the observed latencies fell, respectively.
1311 clocksource=str Use the given clocksource as the base of timing. The
1312 supported options are:
1314 gettimeofday gettimeofday(2)
1316 clock_gettime clock_gettime(2)
1318 cpu Internal CPU clock source
1320 cpu is the preferred clocksource if it is reliable, as it
1321 is very fast (and fio is heavy on time calls). Fio will
1322 automatically use this clocksource if it's supported and
1323 considered reliable on the system it is running on, unless
1324 another clocksource is specifically set. For x86/x86-64 CPUs,
1325 this means supporting TSC Invariant.
1327 gtod_reduce=bool Enable all of the gettimeofday() reducing options
1328 (disable_clat, disable_slat, disable_bw) plus reduce
1329 precision of the timeout somewhat to really shrink
1330 the gettimeofday() call count. With this option enabled,
1331 we only do about 0.4% of the gtod() calls we would have
1332 done if all time keeping was enabled.
1334 gtod_cpu=int Sometimes it's cheaper to dedicate a single thread of
1335 execution to just getting the current time. Fio (and
1336 databases, for instance) are very intensive on gettimeofday()
1337 calls. With this option, you can set one CPU aside for
1338 doing nothing but logging current time to a shared memory
1339 location. Then the other threads/processes that run IO
1340 workloads need only copy that segment, instead of entering
1341 the kernel with a gettimeofday() call. The CPU set aside
1342 for doing these time calls will be excluded from other
1343 uses. Fio will manually clear it from the CPU mask of other
1346 continue_on_error=str Normally fio will exit the job on the first observed
1347 failure. If this option is set, fio will continue the job when
1348 there is a 'non-fatal error' (EIO or EILSEQ) until the runtime
1349 is exceeded or the I/O size specified is completed. If this
1350 option is used, there are two more stats that are appended,
1351 the total error count and the first error. The error field
1352 given in the stats is the first error that was hit during the
1355 The allowed values are:
1357 none Exit on any IO or verify errors.
1359 read Continue on read errors, exit on all others.
1361 write Continue on write errors, exit on all others.
1363 io Continue on any IO error, exit on all others.
1365 verify Continue on verify errors, exit on all others.
1367 all Continue on all errors.
1369 0 Backward-compatible alias for 'none'.
1371 1 Backward-compatible alias for 'all'.
1373 ignore_error=str Sometimes you want to ignore some errors during test
1374 in that case you can specify error list for each error type.
1375 ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST
1376 errors for given error type is separated with ':'. Error
1377 may be symbol ('ENOSPC', 'ENOMEM') or integer.
1379 ignore_error=EAGAIN,ENOSPC:122
1380 This option will ignore EAGAIN from READ, and ENOSPC and
1381 122(EDQUOT) from WRITE.
1383 error_dump=bool If set dump every error even if it is non fatal, true
1384 by default. If disabled only fatal error will be dumped
1386 cgroup=str Add job to this control group. If it doesn't exist, it will
1387 be created. The system must have a mounted cgroup blkio
1388 mount point for this to work. If your system doesn't have it
1389 mounted, you can do so with:
1391 # mount -t cgroup -o blkio none /cgroup
1393 cgroup_weight=int Set the weight of the cgroup to this value. See
1394 the documentation that comes with the kernel, allowed values
1395 are in the range of 100..1000.
1397 cgroup_nodelete=bool Normally fio will delete the cgroups it has created after
1398 the job completion. To override this behavior and to leave
1399 cgroups around after the job completion, set cgroup_nodelete=1.
1400 This can be useful if one wants to inspect various cgroup
1401 files after job completion. Default: false
1403 uid=int Instead of running as the invoking user, set the user ID to
1404 this value before the thread/process does any work.
1406 gid=int Set group ID, see uid.
1408 flow_id=int The ID of the flow. If not specified, it defaults to being a
1409 global flow. See flow.
1411 flow=int Weight in token-based flow control. If this value is used, then
1412 there is a 'flow counter' which is used to regulate the
1413 proportion of activity between two or more jobs. fio attempts
1414 to keep this flow counter near zero. The 'flow' parameter
1415 stands for how much should be added or subtracted to the flow
1416 counter on each iteration of the main I/O loop. That is, if
1417 one job has flow=8 and another job has flow=-1, then there
1418 will be a roughly 1:8 ratio in how much one runs vs the other.
1420 flow_watermark=int The maximum value that the absolute value of the flow
1421 counter is allowed to reach before the job must wait for a
1422 lower value of the counter.
1424 flow_sleep=int The period of time, in microseconds, to wait after the flow
1425 watermark has been exceeded before retrying operations
1427 In addition, there are some parameters which are only valid when a specific
1428 ioengine is in use. These are used identically to normal parameters, with the
1429 caveat that when used on the command line, they must come after the ioengine
1430 that defines them is selected.
1432 [libaio] userspace_reap Normally, with the libaio engine in use, fio will use
1433 the io_getevents system call to reap newly returned events.
1434 With this flag turned on, the AIO ring will be read directly
1435 from user-space to reap events. The reaping mode is only
1436 enabled when polling for a minimum of 0 events (eg when
1437 iodepth_batch_complete=0).
1439 [cpu] cpuload=int Attempt to use the specified percentage of CPU cycles.
1441 [cpu] cpuchunks=int Split the load into cycles of the given time. In
1444 [netsplice] hostname=str
1445 [net] hostname=str The host name or IP address to use for TCP or UDP based IO.
1446 If the job is a TCP listener or UDP reader, the hostname is not
1447 used and must be omitted unless it is a valid UDP multicast
1450 [netsplice] port=int
1451 [net] port=int The TCP or UDP port to bind to or connect to.
1453 [netsplice] interface=str
1454 [net] interface=str The IP address of the network interface used to send or
1455 receive UDP multicast
1458 [net] ttl=int Time-to-live value for outgoing UDP multicast packets.
1461 [netsplice] nodelay=bool
1462 [net] nodelay=bool Set TCP_NODELAY on TCP connections.
1464 [netsplice] protocol=str
1465 [netsplice] proto=str
1467 [net] proto=str The network protocol to use. Accepted values are:
1469 tcp Transmission control protocol
1470 udp User datagram protocol
1471 unix UNIX domain socket
1473 When the protocol is TCP or UDP, the port must also be given,
1474 as well as the hostname if the job is a TCP listener or UDP
1475 reader. For unix sockets, the normal filename option should be
1476 used and the port is invalid.
1478 [net] listen For TCP network connections, tell fio to listen for incoming
1479 connections rather than initiating an outgoing connection. The
1480 hostname must be omitted if this option is used.
1481 [net] pingpong Normaly a network writer will just continue writing data, and
1482 a network reader will just consume packages. If pingpong=1
1483 is set, a writer will send its normal payload to the reader,
1484 then wait for the reader to send the same payload back. This
1485 allows fio to measure network latencies. The submission
1486 and completion latencies then measure local time spent
1487 sending or receiving, and the completion latency measures
1488 how long it took for the other end to receive and send back.
1489 For UDP multicast traffic pingpong=1 should only be set for a
1490 single reader when multiple readers are listening to the same
1493 [e4defrag] donorname=str
1494 File will be used as a block donor(swap extents between files)
1495 [e4defrag] inplace=int
1496 Configure donor file blocks allocation strategy
1497 0(default): Preallocate donor's file on init
1498 1 : allocate space immidietly inside defragment event,
1499 and free right after event
1503 6.0 Interpreting the output
1504 ---------------------------
1506 fio spits out a lot of output. While running, fio will display the
1507 status of the jobs created. An example of that would be:
1509 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1511 The characters inside the square brackets denote the current status of
1512 each thread. The possible values (in typical life cycle order) are:
1516 P Thread setup, but not started.
1518 I Thread initialized, waiting or generating necessary data.
1519 p Thread running pre-reading file(s).
1520 R Running, doing sequential reads.
1521 r Running, doing random reads.
1522 W Running, doing sequential writes.
1523 w Running, doing random writes.
1524 M Running, doing mixed sequential reads/writes.
1525 m Running, doing mixed random reads/writes.
1526 F Running, currently waiting for fsync()
1527 V Running, doing verification of written data.
1528 E Thread exited, not reaped by main thread yet.
1530 X Thread reaped, exited with an error.
1531 K Thread reaped, exited due to signal.
1533 The other values are fairly self explanatory - number of threads
1534 currently running and doing io, rate of io since last check (read speed
1535 listed first, then write speed), and the estimated completion percentage
1536 and time for the running group. It's impossible to estimate runtime of
1537 the following groups (if any). Note that the string is displayed in order,
1538 so it's possible to tell which of the jobs are currently doing what. The
1539 first character is the first job defined in the job file, and so forth.
1541 When fio is done (or interrupted by ctrl-c), it will show the data for
1542 each thread, group of threads, and disks in that order. For each data
1543 direction, the output looks like:
1545 Client1 (g=0): err= 0:
1546 write: io= 32MB, bw= 666KB/s, iops=89 , runt= 50320msec
1547 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
1548 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
1549 bw (KB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
1550 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
1551 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
1552 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1553 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1554 issued r/w: total=0/32768, short=0/0
1555 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
1556 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
1558 The client number is printed, along with the group id and error of that
1559 thread. Below is the io statistics, here for writes. In the order listed,
1562 io= Number of megabytes io performed
1563 bw= Average bandwidth rate
1564 iops= Average IOs performed per second
1565 runt= The runtime of that thread
1566 slat= Submission latency (avg being the average, stdev being the
1567 standard deviation). This is the time it took to submit
1568 the io. For sync io, the slat is really the completion
1569 latency, since queue/complete is one operation there. This
1570 value can be in milliseconds or microseconds, fio will choose
1571 the most appropriate base and print that. In the example
1572 above, milliseconds is the best scale. Note: in --minimal mode
1573 latencies are always expressed in microseconds.
1574 clat= Completion latency. Same names as slat, this denotes the
1575 time from submission to completion of the io pieces. For
1576 sync io, clat will usually be equal (or very close) to 0,
1577 as the time from submit to complete is basically just
1578 CPU time (io has already been done, see slat explanation).
1579 bw= Bandwidth. Same names as the xlat stats, but also includes
1580 an approximate percentage of total aggregate bandwidth
1581 this thread received in this group. This last value is
1582 only really useful if the threads in this group are on the
1583 same disk, since they are then competing for disk access.
1584 cpu= CPU usage. User and system time, along with the number
1585 of context switches this thread went through, usage of
1586 system and user time, and finally the number of major
1587 and minor page faults.
1588 IO depths= The distribution of io depths over the job life time. The
1589 numbers are divided into powers of 2, so for example the
1590 16= entries includes depths up to that value but higher
1591 than the previous entry. In other words, it covers the
1592 range from 16 to 31.
1593 IO submit= How many pieces of IO were submitting in a single submit
1594 call. Each entry denotes that amount and below, until
1595 the previous entry - eg, 8=100% mean that we submitted
1596 anywhere in between 5-8 ios per submit call.
1597 IO complete= Like the above submit number, but for completions instead.
1598 IO issued= The number of read/write requests issued, and how many
1600 IO latencies= The distribution of IO completion latencies. This is the
1601 time from when IO leaves fio and when it gets completed.
1602 The numbers follow the same pattern as the IO depths,
1603 meaning that 2=1.6% means that 1.6% of the IO completed
1604 within 2 msecs, 20=12.8% means that 12.8% of the IO
1605 took more than 10 msecs, but less than (or equal to) 20 msecs.
1607 After each client has been listed, the group statistics are printed. They
1608 will look like this:
1610 Run status group 0 (all jobs):
1611 READ: io=64MB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
1612 WRITE: io=64MB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
1614 For each data direction, it prints:
1616 io= Number of megabytes io performed.
1617 aggrb= Aggregate bandwidth of threads in this group.
1618 minb= The minimum average bandwidth a thread saw.
1619 maxb= The maximum average bandwidth a thread saw.
1620 mint= The smallest runtime of the threads in that group.
1621 maxt= The longest runtime of the threads in that group.
1623 And finally, the disk statistics are printed. They will look like this:
1625 Disk stats (read/write):
1626 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
1628 Each value is printed for both reads and writes, with reads first. The
1631 ios= Number of ios performed by all groups.
1632 merge= Number of merges io the io scheduler.
1633 ticks= Number of ticks we kept the disk busy.
1634 io_queue= Total time spent in the disk queue.
1635 util= The disk utilization. A value of 100% means we kept the disk
1636 busy constantly, 50% would be a disk idling half of the time.
1638 It is also possible to get fio to dump the current output while it is
1639 running, without terminating the job. To do that, send fio the USR1 signal.
1640 You can also get regularly timed dumps by using the --status-interval
1641 parameter, or by creating a file in /tmp named fio-dump-status. If fio
1642 sees this file, it will unlink it and dump the current output status.
1648 For scripted usage where you typically want to generate tables or graphs
1649 of the results, fio can output the results in a semicolon separated format.
1650 The format is one long line of values, such as:
1652 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%
1653 A description of this job goes here.
1655 The job description (if provided) follows on a second line.
1657 To enable terse output, use the --minimal command line option. The first
1658 value is the version of the terse output format. If the output has to
1659 be changed for some reason, this number will be incremented by 1 to
1660 signify that change.
1662 Split up, the format is as follows:
1664 terse version, fio version, jobname, groupid, error
1666 Total IO (KB), bandwidth (KB/sec), IOPS, runtime (msec)
1667 Submission latency: min, max, mean, deviation (usec)
1668 Completion latency: min, max, mean, deviation (usec)
1669 Completion latency percentiles: 20 fields (see below)
1670 Total latency: min, max, mean, deviation (usec)
1671 Bw (KB/s): min, max, aggregate percentage of total, mean, deviation
1673 Total IO (KB), bandwidth (KB/sec), IOPS, runtime (msec)
1674 Submission latency: min, max, mean, deviation (usec)
1675 Completion latency: min, max, mean, deviation (usec)
1676 Completion latency percentiles: 20 fields (see below)
1677 Total latency: min, max, mean, deviation (usec)
1678 Bw (KB/s): min, max, aggregate percentage of total, mean, deviation
1679 CPU usage: user, system, context switches, major faults, minor faults
1680 IO depths: <=1, 2, 4, 8, 16, 32, >=64
1681 IO latencies microseconds: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1682 IO latencies milliseconds: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1683 Disk utilization: Disk name, Read ios, write ios,
1684 Read merges, write merges,
1685 Read ticks, write ticks,
1686 Time spent in queue, disk utilization percentage
1687 Additional Info (dependent on continue_on_error, default off): total # errors, first error code
1689 Additional Info (dependent on description being set): Text description
1691 Completion latency percentiles can be a grouping of up to 20 sets, so
1692 for the terse output fio writes all of them. Each field will look like this:
1696 which is the Xth percentile, and the usec latency associated with it.
1698 For disk utilization, all disks used by fio are shown. So for each disk
1699 there will be a disk utilization section.
1702 8.0 Trace file format
1703 ---------------------
1704 There are two trace file format that you can encounter. The older (v1) format
1705 is unsupported since version 1.20-rc3 (March 2008). It will still be described
1706 below in case that you get an old trace and want to understand it.
1708 In any case the trace is a simple text file with a single action per line.
1711 8.1 Trace file format v1
1712 ------------------------
1713 Each line represents a single io action in the following format:
1717 where rw=0/1 for read/write, and the offset and length entries being in bytes.
1719 This format is not supported in Fio versions => 1.20-rc3.
1722 8.2 Trace file format v2
1723 ------------------------
1724 The second version of the trace file format was added in Fio version 1.17.
1725 It allows to access more then one file per trace and has a bigger set of
1726 possible file actions.
1728 The first line of the trace file has to be:
1732 Following this can be lines in two different formats, which are described below.
1734 The file management format:
1738 The filename is given as an absolute path. The action can be one of these:
1740 add Add the given filename to the trace
1741 open Open the file with the given filename. The filename has to have
1742 been added with the add action before.
1743 close Close the file with the given filename. The file has to have been
1747 The file io action format:
1749 filename action offset length
1751 The filename is given as an absolute path, and has to have been added and opened
1752 before it can be used with this format. The offset and length are given in
1753 bytes. The action can be one of these:
1755 wait Wait for 'offset' microseconds. Everything below 100 is discarded.
1756 read Read 'length' bytes beginning from 'offset'
1757 write Write 'length' bytes beginning from 'offset'
1758 sync fsync() the file
1759 datasync fdatasync() the file
1760 trim trim the given file from the given 'offset' for 'length' bytes
1763 9.0 CPU idleness profiling
1764 --------------------------
1765 In some cases, we want to understand CPU overhead in a test. For example,
1766 we test patches for the specific goodness of whether they reduce CPU usage.
1767 fio implements a balloon approach to create a thread per CPU that runs at
1768 idle priority, meaning that it only runs when nobody else needs the cpu.
1769 By measuring the amount of work completed by the thread, idleness of each
1770 CPU can be derived accordingly.
1772 An unit work is defined as touching a full page of unsigned characters. Mean
1773 and standard deviation of time to complete an unit work is reported in "unit
1774 work" section. Options can be chosen to report detailed percpu idleness or
1775 overall system idleness by aggregating percpu stats.
projects / fio.git / blob