8 5. Detailed list of parameters
13 1.0 Overview and history
14 ------------------------
15 fio was originally written to save me the hassle of writing special test
16 case programs when I wanted to test a specific workload, either for
17 performance reasons or to find/reproduce a bug. The process of writing
18 such a test app can be tiresome, especially if you have to do it often.
19 Hence I needed a tool that would be able to simulate a given io workload
20 without resorting to writing a tailored test case again and again.
22 A test work load is difficult to define, though. There can be any number
23 of processes or threads involved, and they can each be using their own
24 way of generating io. You could have someone dirtying large amounts of
25 memory in an memory mapped file, or maybe several threads issuing
26 reads using asynchronous io. fio needed to be flexible enough to
27 simulate both of these cases, and many more.
31 The first step in getting fio to simulate a desired io workload, is
32 writing a job file describing that specific setup. A job file may contain
33 any number of threads and/or files - the typical contents of the job file
34 is a global section defining shared parameters, and one or more job
35 sections describing the jobs involved. When run, fio parses this file
36 and sets everything up as described. If we break down a job from top to
37 bottom, it contains the following basic parameters:
39 IO type Defines the io pattern issued to the file(s).
40 We may only be reading sequentially from this
41 file(s), or we may be writing randomly. Or even
42 mixing reads and writes, sequentially or randomly.
44 Block size In how large chunks are we issuing io? This may be
45 a single value, or it may describe a range of
48 IO size How much data are we going to be reading/writing.
50 IO engine How do we issue io? We could be memory mapping the
51 file, we could be using regular read/write, we
52 could be using splice, async io, syslet, or even
55 IO depth If the io engine is async, how large a queuing
56 depth do we want to maintain?
58 IO type Should we be doing buffered io, or direct/raw io?
60 Num files How many files are we spreading the workload over.
62 Num threads How many threads or processes should we spread
65 The above are the basic parameters defined for a workload, in addition
66 there's a multitude of parameters that modify other aspects of how this
72 See the README file for command line parameters, there are only a few
75 Running fio is normally the easiest part - you just give it the job file
76 (or job files) as parameters:
80 and it will start doing what the job_file tells it to do. You can give
81 more than one job file on the command line, fio will serialize the running
82 of those files. Internally that is the same as using the 'stonewall'
83 parameter described the the parameter section.
85 If the job file contains only one job, you may as well just give the
86 parameters on the command line. The command line parameters are identical
87 to the job parameters, with a few extra that control global parameters
88 (see README). For example, for the job file parameter iodepth=2, the
89 mirror command line option would be --iodepth 2 or --iodepth=2. You can
90 also use the command line for giving more than one job entry. For each
91 --name option that fio sees, it will start a new job with that name.
92 Command line entries following a --name entry will apply to that job,
93 until there are no more entries or a new --name entry is seen. This is
94 similar to the job file options, where each option applies to the current
95 job until a new [] job entry is seen.
97 fio does not need to run as root, except if the files or devices specified
98 in the job section requires that. Some other options may also be restricted,
99 such as memory locking, io scheduler switching, and decreasing the nice value.
104 As previously described, fio accepts one or more job files describing
105 what it is supposed to do. The job file format is the classic ini file,
106 where the names enclosed in [] brackets define the job name. You are free
107 to use any ascii name you want, except 'global' which has special meaning.
108 A global section sets defaults for the jobs described in that file. A job
109 may override a global section parameter, and a job file may even have
110 several global sections if so desired. A job is only affected by a global
111 section residing above it. If the first character in a line is a ';' or a
112 '#', the entire line is discarded as a comment.
114 So let's look at a really simple job file that defines two processes, each
115 randomly reading from a 128MB file.
117 ; -- start job file --
128 As you can see, the job file sections themselves are empty as all the
129 described parameters are shared. As no filename= option is given, fio
130 makes up a filename for each of the jobs as it sees fit. On the command
131 line, this job would look as follows:
133 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
136 Let's look at an example that has a number of processes writing randomly
139 ; -- start job file --
151 Here we have no global section, as we only have one job defined anyway.
152 We want to use async io here, with a depth of 4 for each file. We also
153 increased the buffer size used to 32KB and define numjobs to 4 to
154 fork 4 identical jobs. The result is 4 processes each randomly writing
155 to their own 64MB file. Instead of using the above job file, you could
156 have given the parameters on the command line. For this case, you would
159 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
161 4.1 Environment variables
162 -------------------------
164 fio also supports environment variable expansion in job files. Any
165 substring of the form "${VARNAME}" as part of an option value (in other
166 words, on the right of the `='), will be expanded to the value of the
167 environment variable called VARNAME. If no such environment variable
168 is defined, or VARNAME is the empty string, the empty string will be
171 As an example, let's look at a sample fio invocation and job file:
173 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
175 ; -- start job file --
182 This will expand to the following equivalent job file at runtime:
184 ; -- start job file --
191 fio ships with a few example job files, you can also look there for
194 4.2 Reserved keywords
195 ---------------------
197 Additionally, fio has a set of reserved keywords that will be replaced
198 internally with the appropriate value. Those keywords are:
200 $pagesize The architecture page size of the running system
201 $mb_memory Megabytes of total memory in the system
202 $ncpus Number of online available CPUs
204 These can be used on the command line or in the job file, and will be
205 automatically substituted with the current system values when the job
206 is run. Simple math is also supported on these keywords, so you can
207 perform actions like:
211 and get that properly expanded to 8 times the size of memory in the
215 5.0 Detailed list of parameters
216 -------------------------------
218 This section describes in details each parameter associated with a job.
219 Some parameters take an option of a given type, such as an integer or
220 a string. The following types are used:
222 str String. This is a sequence of alpha characters.
223 time Integer with possible time suffix. In seconds unless otherwise
224 specified, use eg 10m for 10 minutes. Accepts s/m/h for seconds,
226 int SI integer. A whole number value, which may contain a suffix
227 describing the base of the number. Accepted suffixes are k/m/g/t/p,
228 meaning kilo, mega, giga, tera, and peta. The suffix is not case
229 sensitive, and you may also include trailing 'b' (eg 'kb' is the same
230 as 'k'). So if you want to specify 4096, you could either write
231 out '4096' or just give 4k. The suffixes signify base 2 values, so
232 1024 is 1k and 1024k is 1m and so on, unless the suffix is explicitly
233 set to a base 10 value using 'kib', 'mib', 'gib', etc. If that is the
234 case, then 1000 is used as the multiplier. This can be handy for
235 disks, since manufacturers generally use base 10 values when listing
236 the capacity of a drive. If the option accepts an upper and lower
237 range, use a colon ':' or minus '-' to separate such values. May also
238 include a prefix to indicate numbers base. If 0x is used, the number
239 is assumed to be hexadecimal. See irange.
240 bool Boolean. Usually parsed as an integer, however only defined for
241 true and false (1 and 0).
242 irange Integer range with suffix. Allows value range to be given, such
243 as 1024-4096. A colon may also be used as the separator, eg
244 1k:4k. If the option allows two sets of ranges, they can be
245 specified with a ',' or '/' delimiter: 1k-4k/8k-32k. Also see
247 float_list A list of floating numbers, separated by a ':' character.
249 With the above in mind, here follows the complete list of fio job
252 name=str ASCII name of the job. This may be used to override the
253 name printed by fio for this job. Otherwise the job
254 name is used. On the command line this parameter has the
255 special purpose of also signaling the start of a new
258 description=str Text description of the job. Doesn't do anything except
259 dump this text description when this job is run. It's
262 directory=str Prefix filenames with this directory. Used to place files
263 in a different location than "./".
265 filename=str Fio normally makes up a filename based on the job name,
266 thread number, and file number. If you want to share
267 files between threads in a job or several jobs, specify
268 a filename for each of them to override the default. If
269 the ioengine used is 'net', the filename is the host, port,
270 and protocol to use in the format of =host,port,protocol.
271 See ioengine=net for more. If the ioengine is file based, you
272 can specify a number of files by separating the names with a
273 ':' colon. So if you wanted a job to open /dev/sda and /dev/sdb
274 as the two working files, you would use
275 filename=/dev/sda:/dev/sdb. On Windows, disk devices are accessed
276 as \\.\PhysicalDrive0 for the first device, \\.\PhysicalDrive1
278 Note: Windows and FreeBSD prevent write access to areas of the disk
279 containing in-use data (e.g. filesystems).
280 If the wanted filename does need to include a colon, then escape that
281 with a '\' character.
282 For instance, if the filename is "/dev/dsk/foo@3,0:c",
283 then you would use filename="/dev/dsk/foo@3,0\:c".
284 '-' is a reserved name, meaning stdin or stdout. Which of the
285 two depends on the read/write direction set.
287 opendir=str Tell fio to recursively add any file it can find in this
288 directory and down the file system tree.
290 lockfile=str Fio defaults to not locking any files before it does
291 IO to them. If a file or file descriptor is shared, fio
292 can serialize IO to that file to make the end result
293 consistent. This is usual for emulating real workloads that
294 share files. The lock modes are:
296 none No locking. The default.
297 exclusive Only one thread/process may do IO,
298 excluding all others.
299 readwrite Read-write locking on the file. Many
300 readers may access the file at the
301 same time, but writes get exclusive
304 The option may be post-fixed with a lock batch number. If
305 set, then each thread/process may do that amount of IOs to
306 the file before giving up the lock. Since lock acquisition is
307 expensive, batching the lock/unlocks will speed up IO.
310 rw=str Type of io pattern. Accepted values are:
312 read Sequential reads
313 write Sequential writes
314 randwrite Random writes
315 randread Random reads
316 rw,readwrite Sequential mixed reads and writes
317 randrw Random mixed reads and writes
319 For the mixed io types, the default is to split them 50/50.
320 For certain types of io the result may still be skewed a bit,
321 since the speed may be different. It is possible to specify
322 a number of IO's to do before getting a new offset, this is
323 one by appending a ':<nr>' to the end of the string given.
324 For a random read, it would look like 'rw=randread:8' for
325 passing in an offset modifier with a value of 8. If the
326 suffix is used with a sequential IO pattern, then the value
327 specified will be added to the generated offset for each IO.
328 For instance, using rw=write:4k will skip 4k for every
329 write. It turns sequential IO into sequential IO with holes.
330 See the 'rw_sequencer' option.
332 rw_sequencer=str If an offset modifier is given by appending a number to
333 the rw=<str> line, then this option controls how that
334 number modifies the IO offset being generated. Accepted
337 sequential Generate sequential offset
338 identical Generate the same offset
340 'sequential' is only useful for random IO, where fio would
341 normally generate a new random offset for every IO. If you
342 append eg 8 to randread, you would get a new random offset for
343 every 8 IO's. The result would be a seek for only every 8
344 IO's, instead of for every IO. Use rw=randread:8 to specify
345 that. As sequential IO is already sequential, setting
346 'sequential' for that would not result in any differences.
347 'identical' behaves in a similar fashion, except it sends
348 the same offset 8 number of times before generating a new
351 kb_base=int The base unit for a kilobyte. The defacto base is 2^10, 1024.
352 Storage manufacturers like to use 10^3 or 1000 as a base
353 ten unit instead, for obvious reasons. Allow values are
354 1024 or 1000, with 1024 being the default.
356 randrepeat=bool For random IO workloads, seed the generator in a predictable
357 way so that results are repeatable across repetitions.
359 use_os_rand=bool Fio can either use the random generator supplied by the OS
360 to generator random offsets, or it can use it's own internal
361 generator (based on Tausworthe). Default is to use the
362 internal generator, which is often of better quality and
365 fallocate=str Whether pre-allocation is performed when laying down files.
368 none Do not pre-allocate space
369 posix Pre-allocate via posix_fallocate()
370 keep Pre-allocate via fallocate() with
371 FALLOC_FL_KEEP_SIZE set
372 0 Backward-compatible alias for 'none'
373 1 Backward-compatible alias for 'posix'
375 May not be available on all supported platforms. 'keep' is only
376 available on Linux.If using ZFS on Solaris this must be set to
377 'none' because ZFS doesn't support it. Default: 'posix'.
379 fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
380 on what IO patterns it is likely to issue. Sometimes you
381 want to test specific IO patterns without telling the
382 kernel about it, in which case you can disable this option.
383 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
384 IO and POSIX_FADV_RANDOM for random IO.
386 size=int The total size of file io for this job. Fio will run until
387 this many bytes has been transferred, unless runtime is
388 limited by other options (such as 'runtime', for instance).
389 Unless specific nrfiles and filesize options are given,
390 fio will divide this size between the available files
391 specified by the job. If not set, fio will use the full
392 size of the given files or devices. If the the files
393 do not exist, size must be given. It is also possible to
394 give size as a percentage between 1 and 100. If size=20%
395 is given, fio will use 20% of the full size of the given
398 filesize=int Individual file sizes. May be a range, in which case fio
399 will select sizes for files at random within the given range
400 and limited to 'size' in total (if that is given). If not
401 given, each created file is the same size.
404 fill_fs=bool Sets size to something really large and waits for ENOSPC (no
405 space left on device) as the terminating condition. Only makes
406 sense with sequential write. For a read workload, the mount
407 point will be filled first then IO started on the result. This
408 option doesn't make sense if operating on a raw device node,
409 since the size of that is already known by the file system.
410 Additionally, writing beyond end-of-device will not return
414 bs=int The block size used for the io units. Defaults to 4k. Values
415 can be given for both read and writes. If a single int is
416 given, it will apply to both. If a second int is specified
417 after a comma, it will apply to writes only. In other words,
418 the format is either bs=read_and_write or bs=read,write.
419 bs=4k,8k will thus use 4k blocks for reads, and 8k blocks
420 for writes. If you only wish to set the write size, you
421 can do so by passing an empty read size - bs=,8k will set
422 8k for writes and leave the read default value.
425 ba=int At what boundary to align random IO offsets. Defaults to
426 the same as 'blocksize' the minimum blocksize given.
427 Minimum alignment is typically 512b for using direct IO,
428 though it usually depends on the hardware block size. This
429 option is mutually exclusive with using a random map for
430 files, so it will turn off that option.
432 blocksize_range=irange
433 bsrange=irange Instead of giving a single block size, specify a range
434 and fio will mix the issued io block sizes. The issued
435 io unit will always be a multiple of the minimum value
436 given (also see bs_unaligned). Applies to both reads and
437 writes, however a second range can be given after a comma.
440 bssplit=str Sometimes you want even finer grained control of the
441 block sizes issued, not just an even split between them.
442 This option allows you to weight various block sizes,
443 so that you are able to define a specific amount of
444 block sizes issued. The format for this option is:
446 bssplit=blocksize/percentage:blocksize/percentage
448 for as many block sizes as needed. So if you want to define
449 a workload that has 50% 64k blocks, 10% 4k blocks, and
450 40% 32k blocks, you would write:
452 bssplit=4k/10:64k/50:32k/40
454 Ordering does not matter. If the percentage is left blank,
455 fio will fill in the remaining values evenly. So a bssplit
456 option like this one:
458 bssplit=4k/50:1k/:32k/
460 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
461 always add up to 100, if bssplit is given a range that adds
462 up to more, it will error out.
464 bssplit also supports giving separate splits to reads and
465 writes. The format is identical to what bs= accepts. You
466 have to separate the read and write parts with a comma. So
467 if you want a workload that has 50% 2k reads and 50% 4k reads,
468 while having 90% 4k writes and 10% 8k writes, you would
471 bssplit=2k/50:4k/50,4k/90,8k/10
474 bs_unaligned If this option is given, any byte size value within bsrange
475 may be used as a block range. This typically wont work with
476 direct IO, as that normally requires sector alignment.
478 zero_buffers If this option is given, fio will init the IO buffers to
479 all zeroes. The default is to fill them with random data.
481 refill_buffers If this option is given, fio will refill the IO buffers
482 on every submit. The default is to only fill it at init
483 time and reuse that data. Only makes sense if zero_buffers
484 isn't specified, naturally. If data verification is enabled,
485 refill_buffers is also automatically enabled.
487 scramble_buffers=bool If refill_buffers is too costly and the target is
488 using data deduplication, then setting this option will
489 slightly modify the IO buffer contents to defeat normal
490 de-dupe attempts. This is not enough to defeat more clever
491 block compression attempts, but it will stop naive dedupe of
492 blocks. Default: true.
494 buffer_compress_percentage=int If this is set, then fio will attempt to
495 provide IO buffer content (on WRITEs) that compress to
496 the specified level. Fio does this by providing a mix of
497 random data and zeroes. Note that this is per block size
498 unit, for file/disk wide compression level that matches
499 this setting, you'll also want to set refill_buffers.
501 buffer_compress_chunk=int See buffer_compress_percentage. This
502 setting allows fio to manage how big the ranges of random
503 data and zeroed data is. Without this set, fio will
504 provide buffer_compress_percentage of blocksize random
505 data, followed by the remaining zeroed. With this set
506 to some chunk size smaller than the block size, fio can
507 alternate random and zeroed data throughout the IO
510 nrfiles=int Number of files to use for this job. Defaults to 1.
512 openfiles=int Number of files to keep open at the same time. Defaults to
513 the same as nrfiles, can be set smaller to limit the number
516 file_service_type=str Defines how fio decides which file from a job to
517 service next. The following types are defined:
519 random Just choose a file at random.
521 roundrobin Round robin over open files. This
524 sequential Finish one file before moving on to
525 the next. Multiple files can still be
526 open depending on 'openfiles'.
528 The string can have a number appended, indicating how
529 often to switch to a new file. So if option random:4 is
530 given, fio will switch to a new random file after 4 ios
533 ioengine=str Defines how the job issues io to the file. The following
536 sync Basic read(2) or write(2) io. lseek(2) is
537 used to position the io location.
539 psync Basic pread(2) or pwrite(2) io.
541 vsync Basic readv(2) or writev(2) IO.
543 libaio Linux native asynchronous io. Note that Linux
544 may only support queued behaviour with
545 non-buffered IO (set direct=1 or buffered=0).
546 This engine defines engine specific options.
548 posixaio glibc posix asynchronous io.
550 solarisaio Solaris native asynchronous io.
552 windowsaio Windows native asynchronous io.
554 mmap File is memory mapped and data copied
555 to/from using memcpy(3).
557 splice splice(2) is used to transfer the data and
558 vmsplice(2) to transfer data from user
561 syslet-rw Use the syslet system calls to make
562 regular read/write async.
564 sg SCSI generic sg v3 io. May either be
565 synchronous using the SG_IO ioctl, or if
566 the target is an sg character device
567 we use read(2) and write(2) for asynchronous
570 null Doesn't transfer any data, just pretends
571 to. This is mainly used to exercise fio
572 itself and for debugging/testing purposes.
574 net Transfer over the network to given host:port.
575 Depending on the protocol used, the hostname,
576 port, listen and filename options are used to
577 specify what sort of connection to make, while
578 the protocol option determines which protocol
580 This engine defines engine specific options.
582 netsplice Like net, but uses splice/vmsplice to
583 map data and send/receive.
584 This engine defines engine specific options.
586 cpuio Doesn't transfer any data, but burns CPU
587 cycles according to the cpuload= and
588 cpucycle= options. Setting cpuload=85
589 will cause that job to do nothing but burn
590 85% of the CPU. In case of SMP machines,
591 use numjobs=<no_of_cpu> to get desired CPU
592 usage, as the cpuload only loads a single
593 CPU at the desired rate.
595 guasi The GUASI IO engine is the Generic Userspace
596 Asyncronous Syscall Interface approach
599 http://www.xmailserver.org/guasi-lib.html
601 for more info on GUASI.
603 rdma The RDMA I/O engine supports both RDMA
604 memory semantics (RDMA_WRITE/RDMA_READ) and
605 channel semantics (Send/Recv) for the
606 InfiniBand, RoCE and iWARP protocols.
608 falloc IO engine that does regular fallocate to
609 simulate data transfer as fio ioengine.
610 DDIR_READ does fallocate(,mode = keep_size,)
611 DDIR_WRITE does fallocate(,mode = 0)
612 DDIR_TRIM does fallocate(,mode = punch_hole)
614 e4defrag IO engine that does regular EXT4_IOC_MOVE_EXT
615 ioctls to simulate defragment activity in
616 request to DDIR_WRITE event
618 external Prefix to specify loading an external
619 IO engine object file. Append the engine
620 filename, eg ioengine=external:/tmp/foo.o
621 to load ioengine foo.o in /tmp.
623 iodepth=int This defines how many io units to keep in flight against
624 the file. The default is 1 for each file defined in this
625 job, can be overridden with a larger value for higher
626 concurrency. Note that increasing iodepth beyond 1 will not
627 affect synchronous ioengines (except for small degress when
628 verify_async is in use). Even async engines may impose OS
629 restrictions causing the desired depth not to be achieved.
630 This may happen on Linux when using libaio and not setting
631 direct=1, since buffered IO is not async on that OS. Keep an
632 eye on the IO depth distribution in the fio output to verify
633 that the achieved depth is as expected. Default: 1.
635 iodepth_batch_submit=int
636 iodepth_batch=int This defines how many pieces of IO to submit at once.
637 It defaults to 1 which means that we submit each IO
638 as soon as it is available, but can be raised to submit
639 bigger batches of IO at the time.
641 iodepth_batch_complete=int This defines how many pieces of IO to retrieve
642 at once. It defaults to 1 which means that we'll ask
643 for a minimum of 1 IO in the retrieval process from
644 the kernel. The IO retrieval will go on until we
645 hit the limit set by iodepth_low. If this variable is
646 set to 0, then fio will always check for completed
647 events before queuing more IO. This helps reduce
648 IO latency, at the cost of more retrieval system calls.
650 iodepth_low=int The low water mark indicating when to start filling
651 the queue again. Defaults to the same as iodepth, meaning
652 that fio will attempt to keep the queue full at all times.
653 If iodepth is set to eg 16 and iodepth_low is set to 4, then
654 after fio has filled the queue of 16 requests, it will let
655 the depth drain down to 4 before starting to fill it again.
657 direct=bool If value is true, use non-buffered io. This is usually
658 O_DIRECT. Note that ZFS on Solaris doesn't support direct io.
659 On Windows the synchronous ioengines don't support direct io.
661 buffered=bool If value is true, use buffered io. This is the opposite
662 of the 'direct' option. Defaults to true.
664 offset=int Start io at the given offset in the file. The data before
665 the given offset will not be touched. This effectively
666 caps the file size at real_size - offset.
668 offset_increment=int If this is provided, then the real offset becomes
669 the offset + offset_increment * thread_number, where the
670 thread number is a counter that starts at 0 and is incremented
671 for each job. This option is useful if there are several jobs
672 which are intended to operate on a file in parallel in disjoint
673 segments, with even spacing between the starting points.
675 fsync=int If writing to a file, issue a sync of the dirty data
676 for every number of blocks given. For example, if you give
677 32 as a parameter, fio will sync the file for every 32
678 writes issued. If fio is using non-buffered io, we may
679 not sync the file. The exception is the sg io engine, which
680 synchronizes the disk cache anyway.
682 fdatasync=int Like fsync= but uses fdatasync() to only sync data and not
684 In FreeBSD and Windows there is no fdatasync(), this falls back to
687 sync_file_range=str:val Use sync_file_range() for every 'val' number of
688 write operations. Fio will track range of writes that
689 have happened since the last sync_file_range() call. 'str'
690 can currently be one or more of:
692 wait_before SYNC_FILE_RANGE_WAIT_BEFORE
693 write SYNC_FILE_RANGE_WRITE
694 wait_after SYNC_FILE_RANGE_WAIT_AFTER
696 So if you do sync_file_range=wait_before,write:8, fio would
697 use SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE for
698 every 8 writes. Also see the sync_file_range(2) man page.
699 This option is Linux specific.
701 overwrite=bool If true, writes to a file will always overwrite existing
702 data. If the file doesn't already exist, it will be
703 created before the write phase begins. If the file exists
704 and is large enough for the specified write phase, nothing
707 end_fsync=bool If true, fsync file contents when a write stage has completed.
709 fsync_on_close=bool If true, fio will fsync() a dirty file on close.
710 This differs from end_fsync in that it will happen on every
711 file close, not just at the end of the job.
713 rwmixread=int How large a percentage of the mix should be reads.
715 rwmixwrite=int How large a percentage of the mix should be writes. If both
716 rwmixread and rwmixwrite is given and the values do not add
717 up to 100%, the latter of the two will be used to override
718 the first. This may interfere with a given rate setting,
719 if fio is asked to limit reads or writes to a certain rate.
720 If that is the case, then the distribution may be skewed.
722 random_distribution=str:float By default, fio will use a completely uniform
723 random distribution when asked to perform random IO. Sometimes
724 it is useful to skew the distribution in specific ways,
725 ensuring that some parts of the data is more hot than others.
726 fio includes the following distribution models:
728 random Uniform random distribution
729 zipf Zipf distribution
730 pareto Pareto distribution
732 When using a zipf or pareto distribution, an input value
733 is also needed to define the access pattern. For zipf, this
734 is the zipf theta. For pareto, it's the pareto power. Fio
735 includes a test program, genzipf, that can be used visualize
736 what the given input values will yield in terms of hit rates.
737 If you wanted to use zipf with a theta of 1.2, you would use
738 random_distribution=zipf:1.2 as the option. If a non-uniform
739 model is used, fio will disable use of the random map.
741 norandommap Normally fio will cover every block of the file when doing
742 random IO. If this option is given, fio will just get a
743 new random offset without looking at past io history. This
744 means that some blocks may not be read or written, and that
745 some blocks may be read/written more than once. This option
746 is mutually exclusive with verify= if and only if multiple
747 blocksizes (via bsrange=) are used, since fio only tracks
748 complete rewrites of blocks.
750 softrandommap=bool See norandommap. If fio runs with the random block map
751 enabled and it fails to allocate the map, if this option is
752 set it will continue without a random block map. As coverage
753 will not be as complete as with random maps, this option is
756 random_generator=str Fio supports the following engines for generating
757 IO offsets for random IO:
759 tausworthe Strong 2^88 cycle random number generator
760 lfsr Linear feedback shift register generator
762 Tausworthe is a strong random number generator, but it
763 requires tracking on the side if we want to ensure that
764 blocks are only read or written once. LFSR guarantees
765 that we never generate the same offset twice, and it's
766 also less computationally expensive. It's not a true
767 random generator, however, though for IO purposes it's
768 typically good enough. LFSR only works with single
769 block sizes, not with workloads that use multiple block
770 sizes. If used with such a workload, fio may read or write
771 some blocks multiple times.
773 nice=int Run the job with the given nice value. See man nice(2).
775 prio=int Set the io priority value of this job. Linux limits us to
776 a positive value between 0 and 7, with 0 being the highest.
779 prioclass=int Set the io priority class. See man ionice(1).
781 thinktime=int Stall the job x microseconds after an io has completed before
782 issuing the next. May be used to simulate processing being
783 done by an application. See thinktime_blocks and
787 Only valid if thinktime is set - pretend to spend CPU time
788 doing something with the data received, before falling back
789 to sleeping for the rest of the period specified by
793 Only valid if thinktime is set - control how many blocks
794 to issue, before waiting 'thinktime' usecs. If not set,
795 defaults to 1 which will make fio wait 'thinktime' usecs
798 rate=int Cap the bandwidth used by this job. The number is in bytes/sec,
799 the normal suffix rules apply. You can use rate=500k to limit
800 reads and writes to 500k each, or you can specify read and
801 writes separately. Using rate=1m,500k would limit reads to
802 1MB/sec and writes to 500KB/sec. Capping only reads or
803 writes can be done with rate=,500k or rate=500k,. The former
804 will only limit writes (to 500KB/sec), the latter will only
807 ratemin=int Tell fio to do whatever it can to maintain at least this
808 bandwidth. Failing to meet this requirement, will cause
809 the job to exit. The same format as rate is used for
810 read vs write separation.
812 rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
813 as rate, just specified independently of bandwidth. If the
814 job is given a block size range instead of a fixed value,
815 the smallest block size is used as the metric. The same format
816 as rate is used for read vs write seperation.
818 rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
819 the job to exit. The same format as rate is used for read vs
822 max_latency=int If set, fio will exit the job if it exceeds this maximum
823 latency. It will exit with an ETIME error.
825 ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
828 cpumask=int Set the CPU affinity of this job. The parameter given is a
829 bitmask of allowed CPU's the job may run on. So if you want
830 the allowed CPUs to be 1 and 5, you would pass the decimal
831 value of (1 << 1 | 1 << 5), or 34. See man
832 sched_setaffinity(2). This may not work on all supported
833 operating systems or kernel versions. This option doesn't
834 work well for a higher CPU count than what you can store in
835 an integer mask, so it can only control cpus 1-32. For
836 boxes with larger CPU counts, use cpus_allowed.
838 cpus_allowed=str Controls the same options as cpumask, but it allows a text
839 setting of the permitted CPUs instead. So to use CPUs 1 and
840 5, you would specify cpus_allowed=1,5. This options also
841 allows a range of CPUs. Say you wanted a binding to CPUs
842 1, 5, and 8-15, you would set cpus_allowed=1,5,8-15.
844 numa_cpu_nodes=str Set this job running on spcified NUMA nodes' CPUs. The
845 arguments allow comma delimited list of cpu numbers,
846 A-B ranges, or 'all'. Note, to enable numa options support,
847 fio must be built on a system with libnuma-dev(el) installed.
849 numa_mem_policy=str Set this job's memory policy and corresponding NUMA
850 nodes. Format of the argements:
852 `mode' is one of the following memory policy:
853 default, prefer, bind, interleave, local
854 For `default' and `local' memory policy, no node is
855 needed to be specified.
856 For `prefer', only one node is allowed.
857 For `bind' and `interleave', it allow comma delimited
858 list of numbers, A-B ranges, or 'all'.
860 startdelay=time Start this job the specified number of seconds after fio
861 has started. Only useful if the job file contains several
862 jobs, and you want to delay starting some jobs to a certain
865 runtime=time Tell fio to terminate processing after the specified number
866 of seconds. It can be quite hard to determine for how long
867 a specified job will run, so this parameter is handy to
868 cap the total runtime to a given time.
870 time_based If set, fio will run for the duration of the runtime
871 specified even if the file(s) are completely read or
872 written. It will simply loop over the same workload
873 as many times as the runtime allows.
875 ramp_time=time If set, fio will run the specified workload for this amount
876 of time before logging any performance numbers. Useful for
877 letting performance settle before logging results, thus
878 minimizing the runtime required for stable results. Note
879 that the ramp_time is considered lead in time for a job,
880 thus it will increase the total runtime if a special timeout
881 or runtime is specified.
883 invalidate=bool Invalidate the buffer/page cache parts for this file prior
884 to starting io. Defaults to true.
886 sync=bool Use sync io for buffered writes. For the majority of the
887 io engines, this means using O_SYNC.
890 mem=str Fio can use various types of memory as the io unit buffer.
891 The allowed values are:
893 malloc Use memory from malloc(3) as the buffers.
895 shm Use shared memory as the buffers. Allocated
898 shmhuge Same as shm, but use huge pages as backing.
900 mmap Use mmap to allocate buffers. May either be
901 anonymous memory, or can be file backed if
902 a filename is given after the option. The
903 format is mem=mmap:/path/to/file.
905 mmaphuge Use a memory mapped huge file as the buffer
906 backing. Append filename after mmaphuge, ala
907 mem=mmaphuge:/hugetlbfs/file
909 The area allocated is a function of the maximum allowed
910 bs size for the job, multiplied by the io depth given. Note
911 that for shmhuge and mmaphuge to work, the system must have
912 free huge pages allocated. This can normally be checked
913 and set by reading/writing /proc/sys/vm/nr_hugepages on a
914 Linux system. Fio assumes a huge page is 4MB in size. So
915 to calculate the number of huge pages you need for a given
916 job file, add up the io depth of all jobs (normally one unless
917 iodepth= is used) and multiply by the maximum bs set. Then
918 divide that number by the huge page size. You can see the
919 size of the huge pages in /proc/meminfo. If no huge pages
920 are allocated by having a non-zero number in nr_hugepages,
921 using mmaphuge or shmhuge will fail. Also see hugepage-size.
923 mmaphuge also needs to have hugetlbfs mounted and the file
924 location should point there. So if it's mounted in /huge,
925 you would use mem=mmaphuge:/huge/somefile.
927 iomem_align=int This indiciates the memory alignment of the IO memory buffers.
928 Note that the given alignment is applied to the first IO unit
929 buffer, if using iodepth the alignment of the following buffers
930 are given by the bs used. In other words, if using a bs that is
931 a multiple of the page sized in the system, all buffers will
932 be aligned to this value. If using a bs that is not page
933 aligned, the alignment of subsequent IO memory buffers is the
934 sum of the iomem_align and bs used.
937 Defines the size of a huge page. Must at least be equal
938 to the system setting, see /proc/meminfo. Defaults to 4MB.
939 Should probably always be a multiple of megabytes, so using
940 hugepage-size=Xm is the preferred way to set this to avoid
941 setting a non-pow-2 bad value.
943 exitall When one job finishes, terminate the rest. The default is
944 to wait for each job to finish, sometimes that is not the
947 bwavgtime=int Average the calculated bandwidth over the given time. Value
948 is specified in milliseconds.
950 iopsavgtime=int Average the calculated IOPS over the given time. Value
951 is specified in milliseconds.
953 create_serialize=bool If true, serialize the file creating for the jobs.
954 This may be handy to avoid interleaving of data
955 files, which may greatly depend on the filesystem
956 used and even the number of processors in the system.
958 create_fsync=bool fsync the data file after creation. This is the
961 create_on_open=bool Don't pre-setup the files for IO, just create open()
962 when it's time to do IO to that file.
964 create_only=bool If true, fio will only run the setup phase of the job.
965 If files need to be laid out or updated on disk, only
966 that will be done. The actual job contents are not
969 pre_read=bool If this is given, files will be pre-read into memory before
970 starting the given IO operation. This will also clear
971 the 'invalidate' flag, since it is pointless to pre-read
972 and then drop the cache. This will only work for IO engines
973 that are seekable, since they allow you to read the same data
974 multiple times. Thus it will not work on eg network or splice
977 unlink=bool Unlink the job files when done. Not the default, as repeated
978 runs of that job would then waste time recreating the file
981 loops=int Run the specified number of iterations of this job. Used
982 to repeat the same workload a given number of times. Defaults
985 do_verify=bool Run the verify phase after a write phase. Only makes sense if
986 verify is set. Defaults to 1.
988 verify=str If writing to a file, fio can verify the file contents
989 after each iteration of the job. The allowed values are:
991 md5 Use an md5 sum of the data area and store
992 it in the header of each block.
994 crc64 Use an experimental crc64 sum of the data
995 area and store it in the header of each
998 crc32c Use a crc32c sum of the data area and store
999 it in the header of each block.
1001 crc32c-intel Use hardware assisted crc32c calcuation
1002 provided on SSE4.2 enabled processors. Falls
1003 back to regular software crc32c, if not
1004 supported by the system.
1006 crc32 Use a crc32 sum of the data area and store
1007 it in the header of each block.
1009 crc16 Use a crc16 sum of the data area and store
1010 it in the header of each block.
1012 crc7 Use a crc7 sum of the data area and store
1013 it in the header of each block.
1015 sha512 Use sha512 as the checksum function.
1017 sha256 Use sha256 as the checksum function.
1019 sha1 Use optimized sha1 as the checksum function.
1021 meta Write extra information about each io
1022 (timestamp, block number etc.). The block
1023 number is verified. See also verify_pattern.
1025 null Only pretend to verify. Useful for testing
1026 internals with ioengine=null, not for much
1029 This option can be used for repeated burn-in tests of a
1030 system to make sure that the written data is also
1031 correctly read back. If the data direction given is
1032 a read or random read, fio will assume that it should
1033 verify a previously written file. If the data direction
1034 includes any form of write, the verify will be of the
1037 verifysort=bool If set, fio will sort written verify blocks when it deems
1038 it faster to read them back in a sorted manner. This is
1039 often the case when overwriting an existing file, since
1040 the blocks are already laid out in the file system. You
1041 can ignore this option unless doing huge amounts of really
1042 fast IO where the red-black tree sorting CPU time becomes
1045 verify_offset=int Swap the verification header with data somewhere else
1046 in the block before writing. Its swapped back before
1049 verify_interval=int Write the verification header at a finer granularity
1050 than the blocksize. It will be written for chunks the
1051 size of header_interval. blocksize should divide this
1054 verify_pattern=str If set, fio will fill the io buffers with this
1055 pattern. Fio defaults to filling with totally random
1056 bytes, but sometimes it's interesting to fill with a known
1057 pattern for io verification purposes. Depending on the
1058 width of the pattern, fio will fill 1/2/3/4 bytes of the
1059 buffer at the time(it can be either a decimal or a hex number).
1060 The verify_pattern if larger than a 32-bit quantity has to
1061 be a hex number that starts with either "0x" or "0X". Use
1064 verify_fatal=bool Normally fio will keep checking the entire contents
1065 before quitting on a block verification failure. If this
1066 option is set, fio will exit the job on the first observed
1069 verify_dump=bool If set, dump the contents of both the original data
1070 block and the data block we read off disk to files. This
1071 allows later analysis to inspect just what kind of data
1072 corruption occurred. Off by default.
1074 verify_async=int Fio will normally verify IO inline from the submitting
1075 thread. This option takes an integer describing how many
1076 async offload threads to create for IO verification instead,
1077 causing fio to offload the duty of verifying IO contents
1078 to one or more separate threads. If using this offload
1079 option, even sync IO engines can benefit from using an
1080 iodepth setting higher than 1, as it allows them to have
1081 IO in flight while verifies are running.
1083 verify_async_cpus=str Tell fio to set the given CPU affinity on the
1084 async IO verification threads. See cpus_allowed for the
1087 verify_backlog=int Fio will normally verify the written contents of a
1088 job that utilizes verify once that job has completed. In
1089 other words, everything is written then everything is read
1090 back and verified. You may want to verify continually
1091 instead for a variety of reasons. Fio stores the meta data
1092 associated with an IO block in memory, so for large
1093 verify workloads, quite a bit of memory would be used up
1094 holding this meta data. If this option is enabled, fio
1095 will write only N blocks before verifying these blocks.
1097 will verify the previously written blocks before continuing
1100 verify_backlog_batch=int Control how many blocks fio will verify
1101 if verify_backlog is set. If not set, will default to
1102 the value of verify_backlog (meaning the entire queue
1103 is read back and verified). If verify_backlog_batch is
1104 less than verify_backlog then not all blocks will be verified,
1105 if verify_backlog_batch is larger than verify_backlog, some
1106 blocks will be verified more than once.
1109 wait_for_previous Wait for preceeding jobs in the job file to exit, before
1110 starting this one. Can be used to insert serialization
1111 points in the job file. A stone wall also implies starting
1112 a new reporting group.
1114 new_group Start a new reporting group. See: group_reporting.
1116 numjobs=int Create the specified number of clones of this job. May be
1117 used to setup a larger number of threads/processes doing
1118 the same thing. Each thread is reported separately; to see
1119 statistics for all clones as a whole, use group_reporting in
1120 conjunction with new_group.
1122 group_reporting It may sometimes be interesting to display statistics for
1123 groups of jobs as a whole instead of for each individual job.
1124 This is especially true if 'numjobs' is used; looking at
1125 individual thread/process output quickly becomes unwieldy.
1126 To see the final report per-group instead of per-job, use
1127 'group_reporting'. Jobs in a file will be part of the same
1128 reporting group, unless if separated by a stonewall, or by
1131 thread fio defaults to forking jobs, however if this option is
1132 given, fio will use pthread_create(3) to create threads
1135 zonesize=int Divide a file into zones of the specified size. See zoneskip.
1137 zoneskip=int Skip the specified number of bytes when zonesize data has
1138 been read. The two zone options can be used to only do
1139 io on zones of a file.
1141 write_iolog=str Write the issued io patterns to the specified file. See
1142 read_iolog. Specify a separate file for each job, otherwise
1143 the iologs will be interspersed and the file may be corrupt.
1145 read_iolog=str Open an iolog with the specified file name and replay the
1146 io patterns it contains. This can be used to store a
1147 workload and replay it sometime later. The iolog given
1148 may also be a blktrace binary file, which allows fio
1149 to replay a workload captured by blktrace. See blktrace
1150 for how to capture such logging data. For blktrace replay,
1151 the file needs to be turned into a blkparse binary data
1152 file first (blkparse <device> -o /dev/null -d file_for_fio.bin).
1154 replay_no_stall=int When replaying I/O with read_iolog the default behavior
1155 is to attempt to respect the time stamps within the log and
1156 replay them with the appropriate delay between IOPS. By
1157 setting this variable fio will not respect the timestamps and
1158 attempt to replay them as fast as possible while still
1159 respecting ordering. The result is the same I/O pattern to a
1160 given device, but different timings.
1162 replay_redirect=str While replaying I/O patterns using read_iolog the
1163 default behavior is to replay the IOPS onto the major/minor
1164 device that each IOP was recorded from. This is sometimes
1165 undesireable because on a different machine those major/minor
1166 numbers can map to a different device. Changing hardware on
1167 the same system can also result in a different major/minor
1168 mapping. Replay_redirect causes all IOPS to be replayed onto
1169 the single specified device regardless of the device it was
1170 recorded from. i.e. replay_redirect=/dev/sdc would cause all
1171 IO in the blktrace to be replayed onto /dev/sdc. This means
1172 multiple devices will be replayed onto a single, if the trace
1173 contains multiple devices. If you want multiple devices to be
1174 replayed concurrently to multiple redirected devices you must
1175 blkparse your trace into separate traces and replay them with
1176 independent fio invocations. Unfortuantely this also breaks
1177 the strict time ordering between multiple device accesses.
1179 write_bw_log=str If given, write a bandwidth log of the jobs in this job
1180 file. Can be used to store data of the bandwidth of the
1181 jobs in their lifetime. The included fio_generate_plots
1182 script uses gnuplot to turn these text files into nice
1183 graphs. See write_lat_log for behaviour of given
1184 filename. For this option, the suffix is _bw.log.
1186 write_lat_log=str Same as write_bw_log, except that this option stores io
1187 submission, completion, and total latencies instead. If no
1188 filename is given with this option, the default filename of
1189 "jobname_type.log" is used. Even if the filename is given,
1190 fio will still append the type of log. So if one specifies
1194 The actual log names will be foo_slat.log, foo_slat.log,
1195 and foo_lat.log. This helps fio_generate_plot fine the logs
1198 write_bw_log=str If given, write an IOPS log of the jobs in this job
1199 file. See write_bw_log.
1201 write_iops_log=str Same as write_bw_log, but writes IOPS. If no filename is
1202 given with this option, the default filename of
1203 "jobname_type.log" is used. Even if the filename is given,
1204 fio will still append the type of log.
1206 log_avg_msec=int By default, fio will log an entry in the iops, latency,
1207 or bw log for every IO that completes. When writing to the
1208 disk log, that can quickly grow to a very large size. Setting
1209 this option makes fio average the each log entry over the
1210 specified period of time, reducing the resolution of the log.
1213 lockmem=int Pin down the specified amount of memory with mlock(2). Can
1214 potentially be used instead of removing memory or booting
1215 with less memory to simulate a smaller amount of memory.
1217 exec_prerun=str Before running this job, issue the command specified
1220 exec_postrun=str After the job completes, issue the command specified
1223 ioscheduler=str Attempt to switch the device hosting the file to the specified
1224 io scheduler before running.
1226 cpuload=int If the job is a CPU cycle eater, attempt to use the specified
1227 percentage of CPU cycles.
1229 cpuchunks=int If the job is a CPU cycle eater, split the load into
1230 cycles of the given time. In microseconds.
1232 disk_util=bool Generate disk utilization statistics, if the platform
1233 supports it. Defaults to on.
1235 disable_lat=bool Disable measurements of total latency numbers. Useful
1236 only for cutting back the number of calls to gettimeofday,
1237 as that does impact performance at really high IOPS rates.
1238 Note that to really get rid of a large amount of these
1239 calls, this option must be used with disable_slat and
1242 disable_clat=bool Disable measurements of completion latency numbers. See
1245 disable_slat=bool Disable measurements of submission latency numbers. See
1248 disable_bw=bool Disable measurements of throughput/bandwidth numbers. See
1251 clat_percentiles=bool Enable the reporting of percentiles of
1252 completion latencies.
1254 percentile_list=float_list Overwrite the default list of percentiles
1255 for completion latencies. Each number is a floating
1256 number in the range (0,100], and the maximum length of
1257 the list is 20. Use ':' to separate the numbers, and
1258 list the numbers in ascending order. For example,
1259 --percentile_list=99.5:99.9 will cause fio to report
1260 the values of completion latency below which 99.5% and
1261 99.9% of the observed latencies fell, respectively.
1263 clocksource=str Use the given clocksource as the base of timing. The
1264 supported options are:
1266 gettimeofday gettimeofday(2)
1268 clock_gettime clock_gettime(2)
1270 cpu Internal CPU clock source
1272 cpu is the preferred clocksource if it is reliable, as it
1273 is very fast (and fio is heavy on time calls). Fio will
1274 automatically use this clocksource if it's supported and
1275 considered reliable on the system it is running on, unless
1276 another clocksource is specifically set. For x86/x86-64 CPUs,
1277 this means supporting TSC Invariant.
1279 gtod_reduce=bool Enable all of the gettimeofday() reducing options
1280 (disable_clat, disable_slat, disable_bw) plus reduce
1281 precision of the timeout somewhat to really shrink
1282 the gettimeofday() call count. With this option enabled,
1283 we only do about 0.4% of the gtod() calls we would have
1284 done if all time keeping was enabled.
1286 gtod_cpu=int Sometimes it's cheaper to dedicate a single thread of
1287 execution to just getting the current time. Fio (and
1288 databases, for instance) are very intensive on gettimeofday()
1289 calls. With this option, you can set one CPU aside for
1290 doing nothing but logging current time to a shared memory
1291 location. Then the other threads/processes that run IO
1292 workloads need only copy that segment, instead of entering
1293 the kernel with a gettimeofday() call. The CPU set aside
1294 for doing these time calls will be excluded from other
1295 uses. Fio will manually clear it from the CPU mask of other
1298 continue_on_error=str Normally fio will exit the job on the first observed
1299 failure. If this option is set, fio will continue the job when
1300 there is a 'non-fatal error' (EIO or EILSEQ) until the runtime
1301 is exceeded or the I/O size specified is completed. If this
1302 option is used, there are two more stats that are appended,
1303 the total error count and the first error. The error field
1304 given in the stats is the first error that was hit during the
1307 The allowed values are:
1309 none Exit on any IO or verify errors.
1311 read Continue on read errors, exit on all others.
1313 write Continue on write errors, exit on all others.
1315 io Continue on any IO error, exit on all others.
1317 verify Continue on verify errors, exit on all others.
1319 all Continue on all errors.
1321 0 Backward-compatible alias for 'none'.
1323 1 Backward-compatible alias for 'all'.
1325 ignore_error=str Sometimes you want to ignore some errors during test
1326 in that case you can specify error list for each error type.
1327 ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST
1328 errors for given error type is separated with ':'. Error
1329 may be symbol ('ENOSPC', 'ENOMEM') or integer.
1331 ignore_error=EAGAIN,ENOSPC:122
1332 This option will ignore EAGAIN from READ, and ENOSPC and
1333 122(EDQUOT) from WRITE.
1335 error_dump=bool If set dump every error even if it is non fatal, true
1336 by default. If disabled only fatal error will be dumped
1338 cgroup=str Add job to this control group. If it doesn't exist, it will
1339 be created. The system must have a mounted cgroup blkio
1340 mount point for this to work. If your system doesn't have it
1341 mounted, you can do so with:
1343 # mount -t cgroup -o blkio none /cgroup
1345 cgroup_weight=int Set the weight of the cgroup to this value. See
1346 the documentation that comes with the kernel, allowed values
1347 are in the range of 100..1000.
1349 cgroup_nodelete=bool Normally fio will delete the cgroups it has created after
1350 the job completion. To override this behavior and to leave
1351 cgroups around after the job completion, set cgroup_nodelete=1.
1352 This can be useful if one wants to inspect various cgroup
1353 files after job completion. Default: false
1355 uid=int Instead of running as the invoking user, set the user ID to
1356 this value before the thread/process does any work.
1358 gid=int Set group ID, see uid.
1360 flow_id=int The ID of the flow. If not specified, it defaults to being a
1361 global flow. See flow.
1363 flow=int Weight in token-based flow control. If this value is used, then
1364 there is a 'flow counter' which is used to regulate the
1365 proportion of activity between two or more jobs. fio attempts
1366 to keep this flow counter near zero. The 'flow' parameter
1367 stands for how much should be added or subtracted to the flow
1368 counter on each iteration of the main I/O loop. That is, if
1369 one job has flow=8 and another job has flow=-1, then there
1370 will be a roughly 1:8 ratio in how much one runs vs the other.
1372 flow_watermark=int The maximum value that the absolute value of the flow
1373 counter is allowed to reach before the job must wait for a
1374 lower value of the counter.
1376 flow_sleep=int The period of time, in microseconds, to wait after the flow
1377 watermark has been exceeded before retrying operations
1379 In addition, there are some parameters which are only valid when a specific
1380 ioengine is in use. These are used identically to normal parameters, with the
1381 caveat that when used on the command line, they must come after the ioengine
1382 that defines them is selected.
1384 [libaio] userspace_reap Normally, with the libaio engine in use, fio will use
1385 the io_getevents system call to reap newly returned events.
1386 With this flag turned on, the AIO ring will be read directly
1387 from user-space to reap events. The reaping mode is only
1388 enabled when polling for a minimum of 0 events (eg when
1389 iodepth_batch_complete=0).
1391 [netsplice] hostname=str
1392 [net] hostname=str The host name or IP address to use for TCP or UDP based IO.
1393 If the job is a TCP listener or UDP reader, the hostname is not
1394 used and must be omitted.
1396 [netsplice] port=int
1397 [net] port=int The TCP or UDP port to bind to or connect to.
1399 [netsplice] protocol=str
1400 [netsplice] proto=str
1402 [net] proto=str The network protocol to use. Accepted values are:
1404 tcp Transmission control protocol
1405 udp User datagram protocol
1406 unix UNIX domain socket
1408 When the protocol is TCP or UDP, the port must also be given,
1409 as well as the hostname if the job is a TCP listener or UDP
1410 reader. For unix sockets, the normal filename option should be
1411 used and the port is invalid.
1413 [net] listen For TCP network connections, tell fio to listen for incoming
1414 connections rather than initiating an outgoing connection. The
1415 hostname must be omitted if this option is used.
1416 [net] pingpong Normal a network writer will just continue writing data, and
1417 a network reader will just consume packages. If pingpong=1
1418 is set, a writer will send its normal payload to the reader,
1419 then wait for the reader to send the same payload back. This
1420 allows fio to measure network latencies. The submission
1421 and completion latencies then measure local time spent
1422 sending or receiving, and the completion latency measures
1423 how long it took for the other end to receive and send back.
1425 [e4defrag] donorname=str
1426 File will be used as a block donor(swap extents between files)
1427 [e4defrag] inplace=int
1428 Configure donor file blocks allocation strategy
1429 0(default): Preallocate donor's file on init
1430 1 : allocate space immidietly inside defragment event,
1431 and free right after event
1435 6.0 Interpreting the output
1436 ---------------------------
1438 fio spits out a lot of output. While running, fio will display the
1439 status of the jobs created. An example of that would be:
1441 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1443 The characters inside the square brackets denote the current status of
1444 each thread. The possible values (in typical life cycle order) are:
1448 P Thread setup, but not started.
1450 I Thread initialized, waiting or generating necessary data.
1451 p Thread running pre-reading file(s).
1452 R Running, doing sequential reads.
1453 r Running, doing random reads.
1454 W Running, doing sequential writes.
1455 w Running, doing random writes.
1456 M Running, doing mixed sequential reads/writes.
1457 m Running, doing mixed random reads/writes.
1458 F Running, currently waiting for fsync()
1459 V Running, doing verification of written data.
1460 E Thread exited, not reaped by main thread yet.
1462 X Thread reaped, exited with an error.
1463 K Thread reaped, exited due to signal.
1465 The other values are fairly self explanatory - number of threads
1466 currently running and doing io, rate of io since last check (read speed
1467 listed first, then write speed), and the estimated completion percentage
1468 and time for the running group. It's impossible to estimate runtime of
1469 the following groups (if any). Note that the string is displayed in order,
1470 so it's possible to tell which of the jobs are currently doing what. The
1471 first character is the first job defined in the job file, and so forth.
1473 When fio is done (or interrupted by ctrl-c), it will show the data for
1474 each thread, group of threads, and disks in that order. For each data
1475 direction, the output looks like:
1477 Client1 (g=0): err= 0:
1478 write: io= 32MB, bw= 666KB/s, iops=89 , runt= 50320msec
1479 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
1480 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
1481 bw (KB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
1482 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
1483 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
1484 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1485 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1486 issued r/w: total=0/32768, short=0/0
1487 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
1488 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
1490 The client number is printed, along with the group id and error of that
1491 thread. Below is the io statistics, here for writes. In the order listed,
1494 io= Number of megabytes io performed
1495 bw= Average bandwidth rate
1496 iops= Average IOs performed per second
1497 runt= The runtime of that thread
1498 slat= Submission latency (avg being the average, stdev being the
1499 standard deviation). This is the time it took to submit
1500 the io. For sync io, the slat is really the completion
1501 latency, since queue/complete is one operation there. This
1502 value can be in milliseconds or microseconds, fio will choose
1503 the most appropriate base and print that. In the example
1504 above, milliseconds is the best scale. Note: in --minimal mode
1505 latencies are always expressed in microseconds.
1506 clat= Completion latency. Same names as slat, this denotes the
1507 time from submission to completion of the io pieces. For
1508 sync io, clat will usually be equal (or very close) to 0,
1509 as the time from submit to complete is basically just
1510 CPU time (io has already been done, see slat explanation).
1511 bw= Bandwidth. Same names as the xlat stats, but also includes
1512 an approximate percentage of total aggregate bandwidth
1513 this thread received in this group. This last value is
1514 only really useful if the threads in this group are on the
1515 same disk, since they are then competing for disk access.
1516 cpu= CPU usage. User and system time, along with the number
1517 of context switches this thread went through, usage of
1518 system and user time, and finally the number of major
1519 and minor page faults.
1520 IO depths= The distribution of io depths over the job life time. The
1521 numbers are divided into powers of 2, so for example the
1522 16= entries includes depths up to that value but higher
1523 than the previous entry. In other words, it covers the
1524 range from 16 to 31.
1525 IO submit= How many pieces of IO were submitting in a single submit
1526 call. Each entry denotes that amount and below, until
1527 the previous entry - eg, 8=100% mean that we submitted
1528 anywhere in between 5-8 ios per submit call.
1529 IO complete= Like the above submit number, but for completions instead.
1530 IO issued= The number of read/write requests issued, and how many
1532 IO latencies= The distribution of IO completion latencies. This is the
1533 time from when IO leaves fio and when it gets completed.
1534 The numbers follow the same pattern as the IO depths,
1535 meaning that 2=1.6% means that 1.6% of the IO completed
1536 within 2 msecs, 20=12.8% means that 12.8% of the IO
1537 took more than 10 msecs, but less than (or equal to) 20 msecs.
1539 After each client has been listed, the group statistics are printed. They
1540 will look like this:
1542 Run status group 0 (all jobs):
1543 READ: io=64MB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
1544 WRITE: io=64MB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
1546 For each data direction, it prints:
1548 io= Number of megabytes io performed.
1549 aggrb= Aggregate bandwidth of threads in this group.
1550 minb= The minimum average bandwidth a thread saw.
1551 maxb= The maximum average bandwidth a thread saw.
1552 mint= The smallest runtime of the threads in that group.
1553 maxt= The longest runtime of the threads in that group.
1555 And finally, the disk statistics are printed. They will look like this:
1557 Disk stats (read/write):
1558 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
1560 Each value is printed for both reads and writes, with reads first. The
1563 ios= Number of ios performed by all groups.
1564 merge= Number of merges io the io scheduler.
1565 ticks= Number of ticks we kept the disk busy.
1566 io_queue= Total time spent in the disk queue.
1567 util= The disk utilization. A value of 100% means we kept the disk
1568 busy constantly, 50% would be a disk idling half of the time.
1570 It is also possible to get fio to dump the current output while it is
1571 running, without terminating the job. To do that, send fio the USR1 signal.
1577 For scripted usage where you typically want to generate tables or graphs
1578 of the results, fio can output the results in a semicolon separated format.
1579 The format is one long line of values, such as:
1581 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%
1582 A description of this job goes here.
1584 The job description (if provided) follows on a second line.
1586 To enable terse output, use the --minimal command line option. The first
1587 value is the version of the terse output format. If the output has to
1588 be changed for some reason, this number will be incremented by 1 to
1589 signify that change.
1591 Split up, the format is as follows:
1593 terse version, fio version, jobname, groupid, error
1595 Total IO (KB), bandwidth (KB/sec), IOPS, runtime (msec)
1596 Submission latency: min, max, mean, deviation (usec)
1597 Completion latency: min, max, mean, deviation (usec)
1598 Completion latency percentiles: 20 fields (see below)
1599 Total latency: min, max, mean, deviation (usec)
1600 Bw (KB/s): min, max, aggregate percentage of total, mean, deviation
1602 Total IO (KB), bandwidth (KB/sec), IOPS, runtime (msec)
1603 Submission latency: min, max, mean, deviation (usec)
1604 Completion latency: min, max, mean, deviation (usec)
1605 Completion latency percentiles: 20 fields (see below)
1606 Total latency: min, max, mean, deviation (usec)
1607 Bw (KB/s): min, max, aggregate percentage of total, mean, deviation
1608 CPU usage: user, system, context switches, major faults, minor faults
1609 IO depths: <=1, 2, 4, 8, 16, 32, >=64
1610 IO latencies microseconds: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1611 IO latencies milliseconds: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1612 Disk utilization: Disk name, Read ios, write ios,
1613 Read merges, write merges,
1614 Read ticks, write ticks,
1615 Time spent in queue, disk utilization percentage
1616 Additional Info (dependant on continue_on_error, default off): total # errors, first error code
1618 Additional Info (dependant on description being set): Text description
1620 Completion latency percentiles can be a grouping of up to 20 sets, so
1621 for the terse output fio writes all of them. Each field will look like this:
1625 which is the Xth percentile, and the usec latency associated with it.
1627 For disk utilization, all disks used by fio are shown. So for each disk
1628 there will be a disk utilization section.
1631 8.0 Trace file format
1632 ---------------------
1633 There are two trace file format that you can encounter. The older (v1) format
1634 is unsupported since version 1.20-rc3 (March 2008). It will still be described
1635 below in case that you get an old trace and want to understand it.
1637 In any case the trace is a simple text file with a single action per line.
1640 8.1 Trace file format v1
1641 ------------------------
1642 Each line represents a single io action in the following format:
1646 where rw=0/1 for read/write, and the offset and length entries being in bytes.
1648 This format is not supported in Fio versions => 1.20-rc3.
1651 8.2 Trace file format v2
1652 ------------------------
1653 The second version of the trace file format was added in Fio version 1.17.
1654 It allows to access more then one file per trace and has a bigger set of
1655 possible file actions.
1657 The first line of the trace file has to be:
1661 Following this can be lines in two different formats, which are described below.
1663 The file management format:
1667 The filename is given as an absolute path. The action can be one of these:
1669 add Add the given filename to the trace
1670 open Open the file with the given filename. The filename has to have
1671 been added with the add action before.
1672 close Close the file with the given filename. The file has to have been
1676 The file io action format:
1678 filename action offset length
1680 The filename is given as an absolute path, and has to have been added and opened
1681 before it can be used with this format. The offset and length are given in
1682 bytes. The action can be one of these:
1684 wait Wait for 'offset' microseconds. Everything below 100 is discarded.
1685 read Read 'length' bytes beginning from 'offset'
1686 write Write 'length' bytes beginning from 'offset'
1687 sync fsync() the file
1688 datasync fdatasync() the file
1689 trim trim the given file from the given 'offset' for 'length' bytes
projects / fio.git / blob