8 5. Detailed list of parameters
13 1.0 Overview and history
14 ------------------------
15 fio was originally written to save me the hassle of writing special test
16 case programs when I wanted to test a specific workload, either for
17 performance reasons or to find/reproduce a bug. The process of writing
18 such a test app can be tiresome, especially if you have to do it often.
19 Hence I needed a tool that would be able to simulate a given io workload
20 without resorting to writing a tailored test case again and again.
22 A test work load is difficult to define, though. There can be any number
23 of processes or threads involved, and they can each be using their own
24 way of generating io. You could have someone dirtying large amounts of
25 memory in an memory mapped file, or maybe several threads issuing
26 reads using asynchronous io. fio needed to be flexible enough to
27 simulate both of these cases, and many more.
31 The first step in getting fio to simulate a desired io workload, is
32 writing a job file describing that specific setup. A job file may contain
33 any number of threads and/or files - the typical contents of the job file
34 is a global section defining shared parameters, and one or more job
35 sections describing the jobs involved. When run, fio parses this file
36 and sets everything up as described. If we break down a job from top to
37 bottom, it contains the following basic parameters:
39 IO type Defines the io pattern issued to the file(s).
40 We may only be reading sequentially from this
41 file(s), or we may be writing randomly. Or even
42 mixing reads and writes, sequentially or randomly.
44 Block size In how large chunks are we issuing io? This may be
45 a single value, or it may describe a range of
48 IO size How much data are we going to be reading/writing.
50 IO engine How do we issue io? We could be memory mapping the
51 file, we could be using regular read/write, we
52 could be using splice, async io, syslet, or even
55 IO depth If the io engine is async, how large a queuing
56 depth do we want to maintain?
58 IO type Should we be doing buffered io, or direct/raw io?
60 Num files How many files are we spreading the workload over.
62 Num threads How many threads or processes should we spread
65 The above are the basic parameters defined for a workload, in addition
66 there's a multitude of parameters that modify other aspects of how this
72 See the README file for command line parameters, there are only a few
75 Running fio is normally the easiest part - you just give it the job file
76 (or job files) as parameters:
80 and it will start doing what the job_file tells it to do. You can give
81 more than one job file on the command line, fio will serialize the running
82 of those files. Internally that is the same as using the 'stonewall'
83 parameter described the the parameter section.
85 If the job file contains only one job, you may as well just give the
86 parameters on the command line. The command line parameters are identical
87 to the job parameters, with a few extra that control global parameters
88 (see README). For example, for the job file parameter iodepth=2, the
89 mirror command line option would be --iodepth 2 or --iodepth=2. You can
90 also use the command line for giving more than one job entry. For each
91 --name option that fio sees, it will start a new job with that name.
92 Command line entries following a --name entry will apply to that job,
93 until there are no more entries or a new --name entry is seen. This is
94 similar to the job file options, where each option applies to the current
95 job until a new [] job entry is seen.
97 fio does not need to run as root, except if the files or devices specified
98 in the job section requires that. Some other options may also be restricted,
99 such as memory locking, io scheduler switching, and decreasing the nice value.
104 As previously described, fio accepts one or more job files describing
105 what it is supposed to do. The job file format is the classic ini file,
106 where the names enclosed in [] brackets define the job name. You are free
107 to use any ascii name you want, except 'global' which has special meaning.
108 A global section sets defaults for the jobs described in that file. A job
109 may override a global section parameter, and a job file may even have
110 several global sections if so desired. A job is only affected by a global
111 section residing above it. If the first character in a line is a ';' or a
112 '#', the entire line is discarded as a comment.
114 So let's look at a really simple job file that defines two processes, each
115 randomly reading from a 128MB file.
117 ; -- start job file --
128 As you can see, the job file sections themselves are empty as all the
129 described parameters are shared. As no filename= option is given, fio
130 makes up a filename for each of the jobs as it sees fit. On the command
131 line, this job would look as follows:
133 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
136 Let's look at an example that has a number of processes writing randomly
139 ; -- start job file --
151 Here we have no global section, as we only have one job defined anyway.
152 We want to use async io here, with a depth of 4 for each file. We also
153 increased the buffer size used to 32KB and define numjobs to 4 to
154 fork 4 identical jobs. The result is 4 processes each randomly writing
155 to their own 64MB file. Instead of using the above job file, you could
156 have given the parameters on the command line. For this case, you would
159 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
161 4.1 Environment variables
162 -------------------------
164 fio also supports environment variable expansion in job files. Any
165 substring of the form "${VARNAME}" as part of an option value (in other
166 words, on the right of the `='), will be expanded to the value of the
167 environment variable called VARNAME. If no such environment variable
168 is defined, or VARNAME is the empty string, the empty string will be
171 As an example, let's look at a sample fio invocation and job file:
173 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
175 ; -- start job file --
182 This will expand to the following equivalent job file at runtime:
184 ; -- start job file --
191 fio ships with a few example job files, you can also look there for
194 4.2 Reserved keywords
195 ---------------------
197 Additionally, fio has a set of reserved keywords that will be replaced
198 internally with the appropriate value. Those keywords are:
200 $pagesize The architecture page size of the running system
201 $mb_memory Megabytes of total memory in the system
202 $ncpus Number of online available CPUs
204 These can be used on the command line or in the job file, and will be
205 automatically substituted with the current system values when the job
206 is run. Simple math is also supported on these keywords, so you can
207 perform actions like:
211 and get that properly expanded to 8 times the size of memory in the
215 5.0 Detailed list of parameters
216 -------------------------------
218 This section describes in details each parameter associated with a job.
219 Some parameters take an option of a given type, such as an integer or
220 a string. The following types are used:
222 str String. This is a sequence of alpha characters.
223 time Integer with possible time suffix. In seconds unless otherwise
224 specified, use eg 10m for 10 minutes. Accepts s/m/h for seconds,
226 int SI integer. A whole number value, which may contain a suffix
227 describing the base of the number. Accepted suffixes are k/m/g/t/p,
228 meaning kilo, mega, giga, tera, and peta. The suffix is not case
229 sensitive, and you may also include trailing 'b' (eg 'kb' is the same
230 as 'k'). So if you want to specify 4096, you could either write
231 out '4096' or just give 4k. The suffixes signify base 2 values, so
232 1024 is 1k and 1024k is 1m and so on, unless the suffix is explicitly
233 set to a base 10 value using 'kib', 'mib', 'gib', etc. If that is the
234 case, then 1000 is used as the multiplier. This can be handy for
235 disks, since manufacturers generally use base 10 values when listing
236 the capacity of a drive. If the option accepts an upper and lower
237 range, use a colon ':' or minus '-' to separate such values. May also
238 include a prefix to indicate numbers base. If 0x is used, the number
239 is assumed to be hexadecimal. See irange.
240 bool Boolean. Usually parsed as an integer, however only defined for
241 true and false (1 and 0).
242 irange Integer range with suffix. Allows value range to be given, such
243 as 1024-4096. A colon may also be used as the separator, eg
244 1k:4k. If the option allows two sets of ranges, they can be
245 specified with a ',' or '/' delimiter: 1k-4k/8k-32k. Also see
248 With the above in mind, here follows the complete list of fio job
251 name=str ASCII name of the job. This may be used to override the
252 name printed by fio for this job. Otherwise the job
253 name is used. On the command line this parameter has the
254 special purpose of also signaling the start of a new
257 description=str Text description of the job. Doesn't do anything except
258 dump this text description when this job is run. It's
261 directory=str Prefix filenames with this directory. Used to place files
262 in a different location than "./".
264 filename=str Fio normally makes up a filename based on the job name,
265 thread number, and file number. If you want to share
266 files between threads in a job or several jobs, specify
267 a filename for each of them to override the default. If
268 the ioengine used is 'net', the filename is the host, port,
269 and protocol to use in the format of =host/port/protocol.
270 See ioengine=net for more. If the ioengine is file based, you
271 can specify a number of files by separating the names with a
272 ':' colon. So if you wanted a job to open /dev/sda and /dev/sdb
273 as the two working files, you would use
274 filename=/dev/sda:/dev/sdb. On Windows, disk devices are accessed
275 as \\.\PhysicalDrive0 for the first device, \\.\PhysicalDrive1
276 for the second etc. If the wanted filename does need to
277 include a colon, then escape that with a '\' character.
278 For instance, if the filename is "/dev/dsk/foo@3,0:c",
279 then you would use filename="/dev/dsk/foo@3,0\:c".
280 '-' is a reserved name, meaning stdin or stdout. Which of the
281 two depends on the read/write direction set.
283 opendir=str Tell fio to recursively add any file it can find in this
284 directory and down the file system tree.
286 lockfile=str Fio defaults to not locking any files before it does
287 IO to them. If a file or file descriptor is shared, fio
288 can serialize IO to that file to make the end result
289 consistent. This is usual for emulating real workloads that
290 share files. The lock modes are:
292 none No locking. The default.
293 exclusive Only one thread/process may do IO,
294 excluding all others.
295 readwrite Read-write locking on the file. Many
296 readers may access the file at the
297 same time, but writes get exclusive
300 The option may be post-fixed with a lock batch number. If
301 set, then each thread/process may do that amount of IOs to
302 the file before giving up the lock. Since lock acquisition is
303 expensive, batching the lock/unlocks will speed up IO.
306 rw=str Type of io pattern. Accepted values are:
308 read Sequential reads
309 write Sequential writes
310 randwrite Random writes
311 randread Random reads
312 rw Sequential mixed reads and writes
313 randrw Random mixed reads and writes
315 For the mixed io types, the default is to split them 50/50.
316 For certain types of io the result may still be skewed a bit,
317 since the speed may be different. It is possible to specify
318 a number of IO's to do before getting a new offset, this is
319 one by appending a ':<nr>' to the end of the string given.
320 For a random read, it would look like 'rw=randread:8' for
321 passing in an offset modifier with a value of 8. See the
322 'rw_sequencer' option.
324 rw_sequencer=str If an offset modifier is given by appending a number to
325 the rw=<str> line, then this option controls how that
326 number modifies the IO offset being generated. Accepted
329 sequential Generate sequential offset
330 identical Generate the same offset
332 'sequential' is only useful for random IO, where fio would
333 normally generate a new random offset for every IO. If you
334 append eg 8 to randread, you would get a new random offset for
335 every 8 IO's. The result would be a seek for only every 8
336 IO's, instead of for every IO. Use rw=randread:8 to specify
337 that. As sequential IO is already sequential, setting
338 'sequential' for that would not result in any differences.
339 'identical' behaves in a similar fashion, except it sends
340 the same offset 8 number of times before generating a new
343 kb_base=int The base unit for a kilobyte. The defacto base is 2^10, 1024.
344 Storage manufacturers like to use 10^3 or 1000 as a base
345 ten unit instead, for obvious reasons. Allow values are
346 1024 or 1000, with 1024 being the default.
348 randrepeat=bool For random IO workloads, seed the generator in a predictable
349 way so that results are repeatable across repetitions.
351 use_os_rand=bool Fio can either use the random generator supplied by the OS
352 to generator random offsets, or it can use it's own internal
353 generator (based on Tausworthe). Default is to use the
354 internal generator, which is often of better quality and
357 fallocate=str Whether pre-allocation is performed when laying down files.
360 none Do not pre-allocate space
361 posix Pre-allocate via posix_fallocate()
362 keep Pre-allocate via fallocate() with
363 FALLOC_FL_KEEP_SIZE set
364 0 Backward-compatible alias for 'none'
365 1 Backward-compatible alias for 'posix'
367 May not be available on all supported platforms. 'keep' is only
368 available on Linux.If using ZFS on Solaris this must be set to
369 'none' because ZFS doesn't support it. Default: 'posix'.
371 fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
372 on what IO patterns it is likely to issue. Sometimes you
373 want to test specific IO patterns without telling the
374 kernel about it, in which case you can disable this option.
375 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
376 IO and POSIX_FADV_RANDOM for random IO.
378 size=int The total size of file io for this job. Fio will run until
379 this many bytes has been transferred, unless runtime is
380 limited by other options (such as 'runtime', for instance).
381 Unless specific nrfiles and filesize options are given,
382 fio will divide this size between the available files
383 specified by the job. If not set, fio will use the full
384 size of the given files or devices. If the the files
385 do not exist, size must be given. It is also possible to
386 give size as a percentage between 1 and 100. If size=20%
387 is given, fio will use 20% of the full size of the given
390 filesize=int Individual file sizes. May be a range, in which case fio
391 will select sizes for files at random within the given range
392 and limited to 'size' in total (if that is given). If not
393 given, each created file is the same size.
396 fill_fs=bool Sets size to something really large and waits for ENOSPC (no
397 space left on device) as the terminating condition. Only makes
398 sense with sequential write. For a read workload, the mount
399 point will be filled first then IO started on the result. This
400 option doesn't make sense if operating on a raw device node,
401 since the size of that is already known by the file system.
402 Additionally, writing beyond end-of-device will not return
406 bs=int The block size used for the io units. Defaults to 4k. Values
407 can be given for both read and writes. If a single int is
408 given, it will apply to both. If a second int is specified
409 after a comma, it will apply to writes only. In other words,
410 the format is either bs=read_and_write or bs=read,write.
411 bs=4k,8k will thus use 4k blocks for reads, and 8k blocks
412 for writes. If you only wish to set the write size, you
413 can do so by passing an empty read size - bs=,8k will set
414 8k for writes and leave the read default value.
417 ba=int At what boundary to align random IO offsets. Defaults to
418 the same as 'blocksize' the minimum blocksize given.
419 Minimum alignment is typically 512b for using direct IO,
420 though it usually depends on the hardware block size. This
421 option is mutually exclusive with using a random map for
422 files, so it will turn off that option.
424 blocksize_range=irange
425 bsrange=irange Instead of giving a single block size, specify a range
426 and fio will mix the issued io block sizes. The issued
427 io unit will always be a multiple of the minimum value
428 given (also see bs_unaligned). Applies to both reads and
429 writes, however a second range can be given after a comma.
432 bssplit=str Sometimes you want even finer grained control of the
433 block sizes issued, not just an even split between them.
434 This option allows you to weight various block sizes,
435 so that you are able to define a specific amount of
436 block sizes issued. The format for this option is:
438 bssplit=blocksize/percentage:blocksize/percentage
440 for as many block sizes as needed. So if you want to define
441 a workload that has 50% 64k blocks, 10% 4k blocks, and
442 40% 32k blocks, you would write:
444 bssplit=4k/10:64k/50:32k/40
446 Ordering does not matter. If the percentage is left blank,
447 fio will fill in the remaining values evenly. So a bssplit
448 option like this one:
450 bssplit=4k/50:1k/:32k/
452 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
453 always add up to 100, if bssplit is given a range that adds
454 up to more, it will error out.
456 bssplit also supports giving separate splits to reads and
457 writes. The format is identical to what bs= accepts. You
458 have to separate the read and write parts with a comma. So
459 if you want a workload that has 50% 2k reads and 50% 4k reads,
460 while having 90% 4k writes and 10% 8k writes, you would
463 bssplit=2k/50:4k/50,4k/90,8k/10
466 bs_unaligned If this option is given, any byte size value within bsrange
467 may be used as a block range. This typically wont work with
468 direct IO, as that normally requires sector alignment.
470 zero_buffers If this option is given, fio will init the IO buffers to
471 all zeroes. The default is to fill them with random data.
473 refill_buffers If this option is given, fio will refill the IO buffers
474 on every submit. The default is to only fill it at init
475 time and reuse that data. Only makes sense if zero_buffers
476 isn't specified, naturally. If data verification is enabled,
477 refill_buffers is also automatically enabled.
479 nrfiles=int Number of files to use for this job. Defaults to 1.
481 openfiles=int Number of files to keep open at the same time. Defaults to
482 the same as nrfiles, can be set smaller to limit the number
485 file_service_type=str Defines how fio decides which file from a job to
486 service next. The following types are defined:
488 random Just choose a file at random.
490 roundrobin Round robin over open files. This
493 sequential Finish one file before moving on to
494 the next. Multiple files can still be
495 open depending on 'openfiles'.
497 The string can have a number appended, indicating how
498 often to switch to a new file. So if option random:4 is
499 given, fio will switch to a new random file after 4 ios
502 ioengine=str Defines how the job issues io to the file. The following
505 sync Basic read(2) or write(2) io. lseek(2) is
506 used to position the io location.
508 psync Basic pread(2) or pwrite(2) io.
510 vsync Basic readv(2) or writev(2) IO.
512 libaio Linux native asynchronous io. Note that Linux
513 may only support queued behaviour with
514 non-buffered IO (set direct=1 or buffered=0).
516 posixaio glibc posix asynchronous io.
518 solarisaio Solaris native asynchronous io.
520 windowsaio Windows native asynchronous io.
522 mmap File is memory mapped and data copied
523 to/from using memcpy(3).
525 splice splice(2) is used to transfer the data and
526 vmsplice(2) to transfer data from user
529 syslet-rw Use the syslet system calls to make
530 regular read/write async.
532 sg SCSI generic sg v3 io. May either be
533 synchronous using the SG_IO ioctl, or if
534 the target is an sg character device
535 we use read(2) and write(2) for asynchronous
538 null Doesn't transfer any data, just pretends
539 to. This is mainly used to exercise fio
540 itself and for debugging/testing purposes.
542 net Transfer over the network to given host:port.
543 'filename' must be set appropriately to
544 filename=host/port/protocol regardless of send
545 or receive, if the latter only the port
546 argument is used. 'host' may be an IP address
547 or hostname, port is the port number to be used,
548 and protocol may be 'udp' or 'tcp'. If no
549 protocol is given, TCP is used.
551 netsplice Like net, but uses splice/vmsplice to
552 map data and send/receive.
554 cpuio Doesn't transfer any data, but burns CPU
555 cycles according to the cpuload= and
556 cpucycle= options. Setting cpuload=85
557 will cause that job to do nothing but burn
558 85% of the CPU. In case of SMP machines,
559 use numjobs=<no_of_cpu> to get desired CPU
560 usage, as the cpuload only loads a single
561 CPU at the desired rate.
563 guasi The GUASI IO engine is the Generic Userspace
564 Asyncronous Syscall Interface approach
567 http://www.xmailserver.org/guasi-lib.html
569 for more info on GUASI.
571 external Prefix to specify loading an external
572 IO engine object file. Append the engine
573 filename, eg ioengine=external:/tmp/foo.o
574 to load ioengine foo.o in /tmp.
576 iodepth=int This defines how many io units to keep in flight against
577 the file. The default is 1 for each file defined in this
578 job, can be overridden with a larger value for higher
579 concurrency. Note that increasing iodepth beyond 1 will not
580 affect synchronous ioengines (except for small degress when
581 verify_async is in use). Even async engines may impose OS
582 restrictions causing the desired depth not to be achieved.
583 This may happen on Linux when using libaio and not setting
584 direct=1, since buffered IO is not async on that OS. Keep an
585 eye on the IO depth distribution in the fio output to verify
586 that the achieved depth is as expected. Default: 1.
588 iodepth_batch_submit=int
589 iodepth_batch=int This defines how many pieces of IO to submit at once.
590 It defaults to 1 which means that we submit each IO
591 as soon as it is available, but can be raised to submit
592 bigger batches of IO at the time.
594 iodepth_batch_complete=int This defines how many pieces of IO to retrieve
595 at once. It defaults to 1 which means that we'll ask
596 for a minimum of 1 IO in the retrieval process from
597 the kernel. The IO retrieval will go on until we
598 hit the limit set by iodepth_low. If this variable is
599 set to 0, then fio will always check for completed
600 events before queuing more IO. This helps reduce
601 IO latency, at the cost of more retrieval system calls.
603 iodepth_low=int The low water mark indicating when to start filling
604 the queue again. Defaults to the same as iodepth, meaning
605 that fio will attempt to keep the queue full at all times.
606 If iodepth is set to eg 16 and iodepth_low is set to 4, then
607 after fio has filled the queue of 16 requests, it will let
608 the depth drain down to 4 before starting to fill it again.
610 direct=bool If value is true, use non-buffered io. This is usually
611 O_DIRECT. Note that ZFS on Solaris doesn't support direct io.
613 buffered=bool If value is true, use buffered io. This is the opposite
614 of the 'direct' option. Defaults to true.
616 offset=int Start io at the given offset in the file. The data before
617 the given offset will not be touched. This effectively
618 caps the file size at real_size - offset.
620 fsync=int If writing to a file, issue a sync of the dirty data
621 for every number of blocks given. For example, if you give
622 32 as a parameter, fio will sync the file for every 32
623 writes issued. If fio is using non-buffered io, we may
624 not sync the file. The exception is the sg io engine, which
625 synchronizes the disk cache anyway.
627 fdatasync=int Like fsync= but uses fdatasync() to only sync data and not
629 In FreeBSD there is no fdatasync(), this falls back to
632 sync_file_range=str:val Use sync_file_range() for every 'val' number of
633 write operations. Fio will track range of writes that
634 have happened since the last sync_file_range() call. 'str'
635 can currently be one or more of:
637 wait_before SYNC_FILE_RANGE_WAIT_BEFORE
638 write SYNC_FILE_RANGE_WRITE
639 wait_after SYNC_FILE_RANGE_WAIT_AFTER
641 So if you do sync_file_range=wait_before,write:8, fio would
642 use SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE for
643 every 8 writes. Also see the sync_file_range(2) man page.
644 This option is Linux specific.
646 overwrite=bool If true, writes to a file will always overwrite existing
647 data. If the file doesn't already exist, it will be
648 created before the write phase begins. If the file exists
649 and is large enough for the specified write phase, nothing
652 end_fsync=bool If true, fsync file contents when the job exits.
654 fsync_on_close=bool If true, fio will fsync() a dirty file on close.
655 This differs from end_fsync in that it will happen on every
656 file close, not just at the end of the job.
658 rwmixread=int How large a percentage of the mix should be reads.
660 rwmixwrite=int How large a percentage of the mix should be writes. If both
661 rwmixread and rwmixwrite is given and the values do not add
662 up to 100%, the latter of the two will be used to override
663 the first. This may interfere with a given rate setting,
664 if fio is asked to limit reads or writes to a certain rate.
665 If that is the case, then the distribution may be skewed.
667 norandommap Normally fio will cover every block of the file when doing
668 random IO. If this option is given, fio will just get a
669 new random offset without looking at past io history. This
670 means that some blocks may not be read or written, and that
671 some blocks may be read/written more than once. This option
672 is mutually exclusive with verify= if and only if multiple
673 blocksizes (via bsrange=) are used, since fio only tracks
674 complete rewrites of blocks.
676 softrandommap See norandommap. If fio runs with the random block map enabled
677 and it fails to allocate the map, if this option is set it
678 will continue without a random block map. As coverage will
679 not be as complete as with random maps, this option is
682 nice=int Run the job with the given nice value. See man nice(2).
684 prio=int Set the io priority value of this job. Linux limits us to
685 a positive value between 0 and 7, with 0 being the highest.
688 prioclass=int Set the io priority class. See man ionice(1).
690 thinktime=int Stall the job x microseconds after an io has completed before
691 issuing the next. May be used to simulate processing being
692 done by an application. See thinktime_blocks and
696 Only valid if thinktime is set - pretend to spend CPU time
697 doing something with the data received, before falling back
698 to sleeping for the rest of the period specified by
702 Only valid if thinktime is set - control how many blocks
703 to issue, before waiting 'thinktime' usecs. If not set,
704 defaults to 1 which will make fio wait 'thinktime' usecs
707 rate=int Cap the bandwidth used by this job. The number is in bytes/sec,
708 the normal suffix rules apply. You can use rate=500k to limit
709 reads and writes to 500k each, or you can specify read and
710 writes separately. Using rate=1m,500k would limit reads to
711 1MB/sec and writes to 500KB/sec. Capping only reads or
712 writes can be done with rate=,500k or rate=500k,. The former
713 will only limit writes (to 500KB/sec), the latter will only
716 ratemin=int Tell fio to do whatever it can to maintain at least this
717 bandwidth. Failing to meet this requirement, will cause
718 the job to exit. The same format as rate is used for
719 read vs write separation.
721 rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
722 as rate, just specified independently of bandwidth. If the
723 job is given a block size range instead of a fixed value,
724 the smallest block size is used as the metric. The same format
725 as rate is used for read vs write seperation.
727 rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
728 the job to exit. The same format as rate is used for read vs
731 ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
734 cpumask=int Set the CPU affinity of this job. The parameter given is a
735 bitmask of allowed CPU's the job may run on. So if you want
736 the allowed CPUs to be 1 and 5, you would pass the decimal
737 value of (1 << 1 | 1 << 5), or 34. See man
738 sched_setaffinity(2). This may not work on all supported
739 operating systems or kernel versions. This option doesn't
740 work well for a higher CPU count than what you can store in
741 an integer mask, so it can only control cpus 1-32. For
742 boxes with larger CPU counts, use cpus_allowed.
744 cpus_allowed=str Controls the same options as cpumask, but it allows a text
745 setting of the permitted CPUs instead. So to use CPUs 1 and
746 5, you would specify cpus_allowed=1,5. This options also
747 allows a range of CPUs. Say you wanted a binding to CPUs
748 1, 5, and 8-15, you would set cpus_allowed=1,5,8-15.
750 startdelay=time Start this job the specified number of seconds after fio
751 has started. Only useful if the job file contains several
752 jobs, and you want to delay starting some jobs to a certain
755 runtime=time Tell fio to terminate processing after the specified number
756 of seconds. It can be quite hard to determine for how long
757 a specified job will run, so this parameter is handy to
758 cap the total runtime to a given time.
760 time_based If set, fio will run for the duration of the runtime
761 specified even if the file(s) are completely read or
762 written. It will simply loop over the same workload
763 as many times as the runtime allows.
765 ramp_time=time If set, fio will run the specified workload for this amount
766 of time before logging any performance numbers. Useful for
767 letting performance settle before logging results, thus
768 minimizing the runtime required for stable results. Note
769 that the ramp_time is considered lead in time for a job,
770 thus it will increase the total runtime if a special timeout
771 or runtime is specified.
773 invalidate=bool Invalidate the buffer/page cache parts for this file prior
774 to starting io. Defaults to true.
776 sync=bool Use sync io for buffered writes. For the majority of the
777 io engines, this means using O_SYNC.
780 mem=str Fio can use various types of memory as the io unit buffer.
781 The allowed values are:
783 malloc Use memory from malloc(3) as the buffers.
785 shm Use shared memory as the buffers. Allocated
788 shmhuge Same as shm, but use huge pages as backing.
790 mmap Use mmap to allocate buffers. May either be
791 anonymous memory, or can be file backed if
792 a filename is given after the option. The
793 format is mem=mmap:/path/to/file.
795 mmaphuge Use a memory mapped huge file as the buffer
796 backing. Append filename after mmaphuge, ala
797 mem=mmaphuge:/hugetlbfs/file
799 The area allocated is a function of the maximum allowed
800 bs size for the job, multiplied by the io depth given. Note
801 that for shmhuge and mmaphuge to work, the system must have
802 free huge pages allocated. This can normally be checked
803 and set by reading/writing /proc/sys/vm/nr_hugepages on a
804 Linux system. Fio assumes a huge page is 4MB in size. So
805 to calculate the number of huge pages you need for a given
806 job file, add up the io depth of all jobs (normally one unless
807 iodepth= is used) and multiply by the maximum bs set. Then
808 divide that number by the huge page size. You can see the
809 size of the huge pages in /proc/meminfo. If no huge pages
810 are allocated by having a non-zero number in nr_hugepages,
811 using mmaphuge or shmhuge will fail. Also see hugepage-size.
813 mmaphuge also needs to have hugetlbfs mounted and the file
814 location should point there. So if it's mounted in /huge,
815 you would use mem=mmaphuge:/huge/somefile.
817 iomem_align=int This indiciates the memory alignment of the IO memory buffers.
818 Note that the given alignment is applied to the first IO unit
819 buffer, if using iodepth the alignment of the following buffers
820 are given by the bs used. In other words, if using a bs that is
821 a multiple of the page sized in the system, all buffers will
822 be aligned to this value. If using a bs that is not page
823 aligned, the alignment of subsequent IO memory buffers is the
824 sum of the iomem_align and bs used.
827 Defines the size of a huge page. Must at least be equal
828 to the system setting, see /proc/meminfo. Defaults to 4MB.
829 Should probably always be a multiple of megabytes, so using
830 hugepage-size=Xm is the preferred way to set this to avoid
831 setting a non-pow-2 bad value.
833 exitall When one job finishes, terminate the rest. The default is
834 to wait for each job to finish, sometimes that is not the
837 bwavgtime=int Average the calculated bandwidth over the given time. Value
838 is specified in milliseconds.
840 create_serialize=bool If true, serialize the file creating for the jobs.
841 This may be handy to avoid interleaving of data
842 files, which may greatly depend on the filesystem
843 used and even the number of processors in the system.
845 create_fsync=bool fsync the data file after creation. This is the
848 create_on_open=bool Don't pre-setup the files for IO, just create open()
849 when it's time to do IO to that file.
851 pre_read=bool If this is given, files will be pre-read into memory before
852 starting the given IO operation. This will also clear
853 the 'invalidate' flag, since it is pointless to pre-read
854 and then drop the cache. This will only work for IO engines
855 that are seekable, since they allow you to read the same data
856 multiple times. Thus it will not work on eg network or splice
859 unlink=bool Unlink the job files when done. Not the default, as repeated
860 runs of that job would then waste time recreating the file
863 loops=int Run the specified number of iterations of this job. Used
864 to repeat the same workload a given number of times. Defaults
867 do_verify=bool Run the verify phase after a write phase. Only makes sense if
868 verify is set. Defaults to 1.
870 verify=str If writing to a file, fio can verify the file contents
871 after each iteration of the job. The allowed values are:
873 md5 Use an md5 sum of the data area and store
874 it in the header of each block.
876 crc64 Use an experimental crc64 sum of the data
877 area and store it in the header of each
880 crc32c Use a crc32c sum of the data area and store
881 it in the header of each block.
883 crc32c-intel Use hardware assisted crc32c calcuation
884 provided on SSE4.2 enabled processors. Falls
885 back to regular software crc32c, if not
886 supported by the system.
888 crc32 Use a crc32 sum of the data area and store
889 it in the header of each block.
891 crc16 Use a crc16 sum of the data area and store
892 it in the header of each block.
894 crc7 Use a crc7 sum of the data area and store
895 it in the header of each block.
897 sha512 Use sha512 as the checksum function.
899 sha256 Use sha256 as the checksum function.
901 sha1 Use optimized sha1 as the checksum function.
903 meta Write extra information about each io
904 (timestamp, block number etc.). The block
905 number is verified. See also verify_pattern.
907 null Only pretend to verify. Useful for testing
908 internals with ioengine=null, not for much
911 This option can be used for repeated burn-in tests of a
912 system to make sure that the written data is also
913 correctly read back. If the data direction given is
914 a read or random read, fio will assume that it should
915 verify a previously written file. If the data direction
916 includes any form of write, the verify will be of the
919 verifysort=bool If set, fio will sort written verify blocks when it deems
920 it faster to read them back in a sorted manner. This is
921 often the case when overwriting an existing file, since
922 the blocks are already laid out in the file system. You
923 can ignore this option unless doing huge amounts of really
924 fast IO where the red-black tree sorting CPU time becomes
927 verify_offset=int Swap the verification header with data somewhere else
928 in the block before writing. Its swapped back before
931 verify_interval=int Write the verification header at a finer granularity
932 than the blocksize. It will be written for chunks the
933 size of header_interval. blocksize should divide this
936 verify_pattern=str If set, fio will fill the io buffers with this
937 pattern. Fio defaults to filling with totally random
938 bytes, but sometimes it's interesting to fill with a known
939 pattern for io verification purposes. Depending on the
940 width of the pattern, fio will fill 1/2/3/4 bytes of the
941 buffer at the time(it can be either a decimal or a hex number).
942 The verify_pattern if larger than a 32-bit quantity has to
943 be a hex number that starts with either "0x" or "0X". Use
946 verify_fatal=bool Normally fio will keep checking the entire contents
947 before quitting on a block verification failure. If this
948 option is set, fio will exit the job on the first observed
951 verify_dump=bool If set, dump the contents of both the original data
952 block and the data block we read off disk to files. This
953 allows later analysis to inspect just what kind of data
954 corruption occurred. On by default.
956 verify_async=int Fio will normally verify IO inline from the submitting
957 thread. This option takes an integer describing how many
958 async offload threads to create for IO verification instead,
959 causing fio to offload the duty of verifying IO contents
960 to one or more separate threads. If using this offload
961 option, even sync IO engines can benefit from using an
962 iodepth setting higher than 1, as it allows them to have
963 IO in flight while verifies are running.
965 verify_async_cpus=str Tell fio to set the given CPU affinity on the
966 async IO verification threads. See cpus_allowed for the
969 verify_backlog=int Fio will normally verify the written contents of a
970 job that utilizes verify once that job has completed. In
971 other words, everything is written then everything is read
972 back and verified. You may want to verify continually
973 instead for a variety of reasons. Fio stores the meta data
974 associated with an IO block in memory, so for large
975 verify workloads, quite a bit of memory would be used up
976 holding this meta data. If this option is enabled, fio
977 will write only N blocks before verifying these blocks.
979 will verify the previously written blocks before continuing
982 verify_backlog_batch=int Control how many blocks fio will verify
983 if verify_backlog is set. If not set, will default to
984 the value of verify_backlog (meaning the entire queue
985 is read back and verified). If verify_backlog_batch is
986 less than verify_backlog then not all blocks will be verified,
987 if verify_backlog_batch is larger than verify_backlog, some
988 blocks will be verified more than once.
990 stonewall Wait for preceeding jobs in the job file to exit, before
991 starting this one. Can be used to insert serialization
992 points in the job file. A stone wall also implies starting
993 a new reporting group.
995 new_group Start a new reporting group. If this option isn't given,
996 jobs in a file will be part of the same reporting group
997 unless separated by a stone wall (or if it's a group
998 by itself, with the numjobs option).
1000 numjobs=int Create the specified number of clones of this job. May be
1001 used to setup a larger number of threads/processes doing
1002 the same thing. We regard that grouping of jobs as a
1005 group_reporting If 'numjobs' is set, it may be interesting to display
1006 statistics for the group as a whole instead of for each
1007 individual job. This is especially true of 'numjobs' is
1008 large, looking at individual thread/process output quickly
1009 becomes unwieldy. If 'group_reporting' is specified, fio
1010 will show the final report per-group instead of per-job.
1012 thread fio defaults to forking jobs, however if this option is
1013 given, fio will use pthread_create(3) to create threads
1016 zonesize=int Divide a file into zones of the specified size. See zoneskip.
1018 zoneskip=int Skip the specified number of bytes when zonesize data has
1019 been read. The two zone options can be used to only do
1020 io on zones of a file.
1022 write_iolog=str Write the issued io patterns to the specified file. See
1023 read_iolog. Specify a separate file for each job, otherwise
1024 the iologs will be interspersed and the file may be corrupt.
1026 read_iolog=str Open an iolog with the specified file name and replay the
1027 io patterns it contains. This can be used to store a
1028 workload and replay it sometime later. The iolog given
1029 may also be a blktrace binary file, which allows fio
1030 to replay a workload captured by blktrace. See blktrace
1031 for how to capture such logging data. For blktrace replay,
1032 the file needs to be turned into a blkparse binary data
1033 file first (blkparse <device> -o /dev/null -d file_for_fio.bin).
1035 replay_no_stall=int When replaying I/O with read_iolog the default behavior
1036 is to attempt to respect the time stamps within the log and
1037 replay them with the appropriate delay between IOPS. By
1038 setting this variable fio will not respect the timestamps and
1039 attempt to replay them as fast as possible while still
1040 respecting ordering. The result is the same I/O pattern to a
1041 given device, but different timings.
1043 replay_redirect=str While replaying I/O patterns using read_iolog the
1044 default behavior is to replay the IOPS onto the major/minor
1045 device that each IOP was recorded from. This is sometimes
1046 undesireable because on a different machine those major/minor
1047 numbers can map to a different device. Changing hardware on
1048 the same system can also result in a different major/minor
1049 mapping. Replay_redirect causes all IOPS to be replayed onto
1050 the single specified device regardless of the device it was
1051 recorded from. i.e. replay_redirect=/dev/sdc would cause all
1052 IO in the blktrace to be replayed onto /dev/sdc. This means
1053 multiple devices will be replayed onto a single, if the trace
1054 contains multiple devices. If you want multiple devices to be
1055 replayed concurrently to multiple redirected devices you must
1056 blkparse your trace into separate traces and replay them with
1057 independent fio invocations. Unfortuantely this also breaks
1058 the strict time ordering between multiple device accesses.
1060 write_bw_log=str If given, write a bandwidth log of the jobs in this job
1061 file. Can be used to store data of the bandwidth of the
1062 jobs in their lifetime. The included fio_generate_plots
1063 script uses gnuplot to turn these text files into nice
1064 graphs. See write_log_log for behaviour of given
1065 filename. For this option, the postfix is _bw.log.
1067 write_lat_log=str Same as write_bw_log, except that this option stores io
1068 submission, completion, and total latencies instead. If no
1069 filename is given with this option, the default filename of
1070 "jobname_type.log" is used. Even if the filename is given,
1071 fio will still append the type of log. So if one specifies
1075 The actual log names will be foo_slat.log, foo_slat.log,
1076 and foo_lat.log. This helps fio_generate_plot fine the logs
1079 lockmem=int Pin down the specified amount of memory with mlock(2). Can
1080 potentially be used instead of removing memory or booting
1081 with less memory to simulate a smaller amount of memory.
1083 exec_prerun=str Before running this job, issue the command specified
1086 exec_postrun=str After the job completes, issue the command specified
1089 ioscheduler=str Attempt to switch the device hosting the file to the specified
1090 io scheduler before running.
1092 cpuload=int If the job is a CPU cycle eater, attempt to use the specified
1093 percentage of CPU cycles.
1095 cpuchunks=int If the job is a CPU cycle eater, split the load into
1096 cycles of the given time. In microseconds.
1098 disk_util=bool Generate disk utilization statistics, if the platform
1099 supports it. Defaults to on.
1101 disable_lat=bool Disable measurements of total latency numbers. Useful
1102 only for cutting back the number of calls to gettimeofday,
1103 as that does impact performance at really high IOPS rates.
1104 Note that to really get rid of a large amount of these
1105 calls, this option must be used with disable_slat and
1108 disable_clat=bool Disable measurements of completion latency numbers. See
1111 disable_slat=bool Disable measurements of submission latency numbers. See
1114 disable_bw=bool Disable measurements of throughput/bandwidth numbers. See
1117 gtod_reduce=bool Enable all of the gettimeofday() reducing options
1118 (disable_clat, disable_slat, disable_bw) plus reduce
1119 precision of the timeout somewhat to really shrink
1120 the gettimeofday() call count. With this option enabled,
1121 we only do about 0.4% of the gtod() calls we would have
1122 done if all time keeping was enabled.
1124 gtod_cpu=int Sometimes it's cheaper to dedicate a single thread of
1125 execution to just getting the current time. Fio (and
1126 databases, for instance) are very intensive on gettimeofday()
1127 calls. With this option, you can set one CPU aside for
1128 doing nothing but logging current time to a shared memory
1129 location. Then the other threads/processes that run IO
1130 workloads need only copy that segment, instead of entering
1131 the kernel with a gettimeofday() call. The CPU set aside
1132 for doing these time calls will be excluded from other
1133 uses. Fio will manually clear it from the CPU mask of other
1136 continue_on_error=bool Normally fio will exit the job on the first observed
1137 failure. If this option is set, fio will continue the job when
1138 there is a 'non-fatal error' (EIO or EILSEQ) until the runtime
1139 is exceeded or the I/O size specified is completed. If this
1140 option is used, there are two more stats that are appended,
1141 the total error count and the first error. The error field
1142 given in the stats is the first error that was hit during the
1145 cgroup=str Add job to this control group. If it doesn't exist, it will
1146 be created. The system must have a mounted cgroup blkio
1147 mount point for this to work. If your system doesn't have it
1148 mounted, you can do so with:
1150 # mount -t cgroup -o blkio none /cgroup
1152 cgroup_weight=int Set the weight of the cgroup to this value. See
1153 the documentation that comes with the kernel, allowed values
1154 are in the range of 100..1000.
1156 cgroup_nodelete=bool Normally fio will delete the cgroups it has created after
1157 the job completion. To override this behavior and to leave
1158 cgroups around after the job completion, set cgroup_nodelete=1.
1159 This can be useful if one wants to inspect various cgroup
1160 files after job completion. Default: false
1162 uid=int Instead of running as the invoking user, set the user ID to
1163 this value before the thread/process does any work.
1165 gid=int Set group ID, see uid.
1167 6.0 Interpreting the output
1168 ---------------------------
1170 fio spits out a lot of output. While running, fio will display the
1171 status of the jobs created. An example of that would be:
1173 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1175 The characters inside the square brackets denote the current status of
1176 each thread. The possible values (in typical life cycle order) are:
1180 P Thread setup, but not started.
1182 I Thread initialized, waiting.
1183 p Thread running pre-reading file(s).
1184 R Running, doing sequential reads.
1185 r Running, doing random reads.
1186 W Running, doing sequential writes.
1187 w Running, doing random writes.
1188 M Running, doing mixed sequential reads/writes.
1189 m Running, doing mixed random reads/writes.
1190 F Running, currently waiting for fsync()
1191 V Running, doing verification of written data.
1192 E Thread exited, not reaped by main thread yet.
1195 The other values are fairly self explanatory - number of threads
1196 currently running and doing io, rate of io since last check (read speed
1197 listed first, then write speed), and the estimated completion percentage
1198 and time for the running group. It's impossible to estimate runtime of
1199 the following groups (if any).
1201 When fio is done (or interrupted by ctrl-c), it will show the data for
1202 each thread, group of threads, and disks in that order. For each data
1203 direction, the output looks like:
1205 Client1 (g=0): err= 0:
1206 write: io= 32MB, bw= 666KB/s, runt= 50320msec
1207 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
1208 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
1209 bw (KB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
1210 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
1211 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
1212 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1213 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1214 issued r/w: total=0/32768, short=0/0
1215 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
1216 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
1218 The client number is printed, along with the group id and error of that
1219 thread. Below is the io statistics, here for writes. In the order listed,
1222 io= Number of megabytes io performed
1223 bw= Average bandwidth rate
1224 runt= The runtime of that thread
1225 slat= Submission latency (avg being the average, stdev being the
1226 standard deviation). This is the time it took to submit
1227 the io. For sync io, the slat is really the completion
1228 latency, since queue/complete is one operation there. This
1229 value can be in milliseconds or microseconds, fio will choose
1230 the most appropriate base and print that. In the example
1231 above, milliseconds is the best scale.
1232 clat= Completion latency. Same names as slat, this denotes the
1233 time from submission to completion of the io pieces. For
1234 sync io, clat will usually be equal (or very close) to 0,
1235 as the time from submit to complete is basically just
1236 CPU time (io has already been done, see slat explanation).
1237 bw= Bandwidth. Same names as the xlat stats, but also includes
1238 an approximate percentage of total aggregate bandwidth
1239 this thread received in this group. This last value is
1240 only really useful if the threads in this group are on the
1241 same disk, since they are then competing for disk access.
1242 cpu= CPU usage. User and system time, along with the number
1243 of context switches this thread went through, usage of
1244 system and user time, and finally the number of major
1245 and minor page faults.
1246 IO depths= The distribution of io depths over the job life time. The
1247 numbers are divided into powers of 2, so for example the
1248 16= entries includes depths up to that value but higher
1249 than the previous entry. In other words, it covers the
1250 range from 16 to 31.
1251 IO submit= How many pieces of IO were submitting in a single submit
1252 call. Each entry denotes that amount and below, until
1253 the previous entry - eg, 8=100% mean that we submitted
1254 anywhere in between 5-8 ios per submit call.
1255 IO complete= Like the above submit number, but for completions instead.
1256 IO issued= The number of read/write requests issued, and how many
1258 IO latencies= The distribution of IO completion latencies. This is the
1259 time from when IO leaves fio and when it gets completed.
1260 The numbers follow the same pattern as the IO depths,
1261 meaning that 2=1.6% means that 1.6% of the IO completed
1262 within 2 msecs, 20=12.8% means that 12.8% of the IO
1263 took more than 10 msecs, but less than (or equal to) 20 msecs.
1265 After each client has been listed, the group statistics are printed. They
1266 will look like this:
1268 Run status group 0 (all jobs):
1269 READ: io=64MB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
1270 WRITE: io=64MB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
1272 For each data direction, it prints:
1274 io= Number of megabytes io performed.
1275 aggrb= Aggregate bandwidth of threads in this group.
1276 minb= The minimum average bandwidth a thread saw.
1277 maxb= The maximum average bandwidth a thread saw.
1278 mint= The smallest runtime of the threads in that group.
1279 maxt= The longest runtime of the threads in that group.
1281 And finally, the disk statistics are printed. They will look like this:
1283 Disk stats (read/write):
1284 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
1286 Each value is printed for both reads and writes, with reads first. The
1289 ios= Number of ios performed by all groups.
1290 merge= Number of merges io the io scheduler.
1291 ticks= Number of ticks we kept the disk busy.
1292 io_queue= Total time spent in the disk queue.
1293 util= The disk utilization. A value of 100% means we kept the disk
1294 busy constantly, 50% would be a disk idling half of the time.
1300 For scripted usage where you typically want to generate tables or graphs
1301 of the results, fio can output the results in a semicolon separated format.
1302 The format is one long line of values, such as:
1304 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%
1305 A description of this job goes here.
1307 The job description (if provided) follows on a second line.
1309 To enable terse output, use the --minimal command line option. The first
1310 value is the version of the terse output format. If the output has to
1311 be changed for some reason, this number will be incremented by 1 to
1312 signify that change.
1314 Split up, the format is as follows:
1316 version, jobname, groupid, error
1318 KB IO, bandwidth (KB/sec), runtime (msec)
1319 Submission latency: min, max, mean, deviation
1320 Completion latency: min, max, mean, deviation
1321 Total latency: min, max, mean, deviation
1322 Bw: min, max, aggregate percentage of total, mean, deviation
1324 KB IO, bandwidth (KB/sec), runtime (msec)
1325 Submission latency: min, max, mean, deviation
1326 Completion latency: min, max, mean, deviation
1327 Total latency: min, max, mean, deviation
1328 Bw: min, max, aggregate percentage of total, mean, deviation
1329 CPU usage: user, system, context switches, major faults, minor faults
1330 IO depths: <=1, 2, 4, 8, 16, 32, >=64
1331 IO latencies microseconds: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1332 IO latencies milliseconds: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1333 Additional Info (dependant on continue_on_error, default off): total # errors, first error code
1335 Additional Info (dependant on description being set): Text description
projects / fio.git / blob