8 5. Detailed list of parameters
12 9. CPU idleness profiling
14 1.0 Overview and history
15 ------------------------
16 fio was originally written to save me the hassle of writing special test
17 case programs when I wanted to test a specific workload, either for
18 performance reasons or to find/reproduce a bug. The process of writing
19 such a test app can be tiresome, especially if you have to do it often.
20 Hence I needed a tool that would be able to simulate a given io workload
21 without resorting to writing a tailored test case again and again.
23 A test work load is difficult to define, though. There can be any number
24 of processes or threads involved, and they can each be using their own
25 way of generating io. You could have someone dirtying large amounts of
26 memory in an memory mapped file, or maybe several threads issuing
27 reads using asynchronous io. fio needed to be flexible enough to
28 simulate both of these cases, and many more.
32 The first step in getting fio to simulate a desired io workload, is
33 writing a job file describing that specific setup. A job file may contain
34 any number of threads and/or files - the typical contents of the job file
35 is a global section defining shared parameters, and one or more job
36 sections describing the jobs involved. When run, fio parses this file
37 and sets everything up as described. If we break down a job from top to
38 bottom, it contains the following basic parameters:
40 IO type Defines the io pattern issued to the file(s).
41 We may only be reading sequentially from this
42 file(s), or we may be writing randomly. Or even
43 mixing reads and writes, sequentially or randomly.
45 Block size In how large chunks are we issuing io? This may be
46 a single value, or it may describe a range of
49 IO size How much data are we going to be reading/writing.
51 IO engine How do we issue io? We could be memory mapping the
52 file, we could be using regular read/write, we
53 could be using splice, async io, syslet, or even
56 IO depth If the io engine is async, how large a queuing
57 depth do we want to maintain?
59 IO type Should we be doing buffered io, or direct/raw io?
61 Num files How many files are we spreading the workload over.
63 Num threads How many threads or processes should we spread
66 The above are the basic parameters defined for a workload, in addition
67 there's a multitude of parameters that modify other aspects of how this
73 See the README file for command line parameters, there are only a few
76 Running fio is normally the easiest part - you just give it the job file
77 (or job files) as parameters:
81 and it will start doing what the job_file tells it to do. You can give
82 more than one job file on the command line, fio will serialize the running
83 of those files. Internally that is the same as using the 'stonewall'
84 parameter described in the parameter section.
86 If the job file contains only one job, you may as well just give the
87 parameters on the command line. The command line parameters are identical
88 to the job parameters, with a few extra that control global parameters
89 (see README). For example, for the job file parameter iodepth=2, the
90 mirror command line option would be --iodepth 2 or --iodepth=2. You can
91 also use the command line for giving more than one job entry. For each
92 --name option that fio sees, it will start a new job with that name.
93 Command line entries following a --name entry will apply to that job,
94 until there are no more entries or a new --name entry is seen. This is
95 similar to the job file options, where each option applies to the current
96 job until a new [] job entry is seen.
98 fio does not need to run as root, except if the files or devices specified
99 in the job section requires that. Some other options may also be restricted,
100 such as memory locking, io scheduler switching, and decreasing the nice value.
105 As previously described, fio accepts one or more job files describing
106 what it is supposed to do. The job file format is the classic ini file,
107 where the names enclosed in [] brackets define the job name. You are free
108 to use any ascii name you want, except 'global' which has special meaning.
109 A global section sets defaults for the jobs described in that file. A job
110 may override a global section parameter, and a job file may even have
111 several global sections if so desired. A job is only affected by a global
112 section residing above it. If the first character in a line is a ';' or a
113 '#', the entire line is discarded as a comment.
115 So let's look at a really simple job file that defines two processes, each
116 randomly reading from a 128MB file.
118 ; -- start job file --
129 As you can see, the job file sections themselves are empty as all the
130 described parameters are shared. As no filename= option is given, fio
131 makes up a filename for each of the jobs as it sees fit. On the command
132 line, this job would look as follows:
134 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
137 Let's look at an example that has a number of processes writing randomly
140 ; -- start job file --
152 Here we have no global section, as we only have one job defined anyway.
153 We want to use async io here, with a depth of 4 for each file. We also
154 increased the buffer size used to 32KB and define numjobs to 4 to
155 fork 4 identical jobs. The result is 4 processes each randomly writing
156 to their own 64MB file. Instead of using the above job file, you could
157 have given the parameters on the command line. For this case, you would
160 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
162 When fio is utilized as a basis of any reasonably large test suite, it might be
163 desirable to share a set of standardized settings across multiple job files.
164 Instead of copy/pasting such settings, any section may pull in an external
165 .fio file with 'include filename' directive, as in the following example:
167 ; -- start job file including.fio --
171 include glob-include.fio
178 include test-include.fio
179 ; -- end job file including.fio --
181 ; -- start job file glob-include.fio --
184 ; -- end job file glob-include.fio --
186 ; -- start job file test-include.fio --
189 ; -- end job file test-include.fio --
191 Settings pulled into a section apply to that section only (except global
192 section). Include directives may be nested in that any included file may
193 contain further include directive(s). Include files may not contain []
197 4.1 Environment variables
198 -------------------------
200 fio also supports environment variable expansion in job files. Any
201 substring of the form "${VARNAME}" as part of an option value (in other
202 words, on the right of the `='), will be expanded to the value of the
203 environment variable called VARNAME. If no such environment variable
204 is defined, or VARNAME is the empty string, the empty string will be
207 As an example, let's look at a sample fio invocation and job file:
209 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
211 ; -- start job file --
218 This will expand to the following equivalent job file at runtime:
220 ; -- start job file --
227 fio ships with a few example job files, you can also look there for
230 4.2 Reserved keywords
231 ---------------------
233 Additionally, fio has a set of reserved keywords that will be replaced
234 internally with the appropriate value. Those keywords are:
236 $pagesize The architecture page size of the running system
237 $mb_memory Megabytes of total memory in the system
238 $ncpus Number of online available CPUs
240 These can be used on the command line or in the job file, and will be
241 automatically substituted with the current system values when the job
242 is run. Simple math is also supported on these keywords, so you can
243 perform actions like:
247 and get that properly expanded to 8 times the size of memory in the
251 5.0 Detailed list of parameters
252 -------------------------------
254 This section describes in details each parameter associated with a job.
255 Some parameters take an option of a given type, such as an integer or
256 a string. Anywhere a numeric value is required, an arithmetic expression
257 may be used, provided it is surrounded by parentheses. Supported operators
267 For time values in expressions, units are microseconds by default. This is
268 different than for time values not in expressions (not enclosed in
269 parentheses). The following types are used:
271 str String. This is a sequence of alpha characters.
272 time Integer with possible time suffix. In seconds unless otherwise
273 specified, use eg 10m for 10 minutes. Accepts s/m/h for seconds,
274 minutes, and hours, and accepts 'ms' (or 'msec') for milliseconds,
275 and 'us' (or 'usec') for microseconds.
276 int SI integer. A whole number value, which may contain a suffix
277 describing the base of the number. Accepted suffixes are k/m/g/t/p,
278 meaning kilo, mega, giga, tera, and peta. The suffix is not case
279 sensitive, and you may also include trailing 'b' (eg 'kb' is the same
280 as 'k'). So if you want to specify 4096, you could either write
281 out '4096' or just give 4k. The suffixes signify base 2 values, so
282 1024 is 1k and 1024k is 1m and so on, unless the suffix is explicitly
283 set to a base 10 value using 'kib', 'mib', 'gib', etc. If that is the
284 case, then 1000 is used as the multiplier. This can be handy for
285 disks, since manufacturers generally use base 10 values when listing
286 the capacity of a drive. If the option accepts an upper and lower
287 range, use a colon ':' or minus '-' to separate such values. May also
288 include a prefix to indicate numbers base. If 0x is used, the number
289 is assumed to be hexadecimal. See irange.
290 bool Boolean. Usually parsed as an integer, however only defined for
291 true and false (1 and 0).
292 irange Integer range with suffix. Allows value range to be given, such
293 as 1024-4096. A colon may also be used as the separator, eg
294 1k:4k. If the option allows two sets of ranges, they can be
295 specified with a ',' or '/' delimiter: 1k-4k/8k-32k. Also see
297 float_list A list of floating numbers, separated by a ':' character.
299 With the above in mind, here follows the complete list of fio job
302 name=str ASCII name of the job. This may be used to override the
303 name printed by fio for this job. Otherwise the job
304 name is used. On the command line this parameter has the
305 special purpose of also signaling the start of a new
308 description=str Text description of the job. Doesn't do anything except
309 dump this text description when this job is run. It's
312 directory=str Prefix filenames with this directory. Used to place files
313 in a different location than "./". See the 'filename' option
314 for escaping certain characters.
316 filename=str Fio normally makes up a filename based on the job name,
317 thread number, and file number. If you want to share
318 files between threads in a job or several jobs, specify
319 a filename for each of them to override the default. If
320 the ioengine used is 'net', the filename is the host, port,
321 and protocol to use in the format of =host,port,protocol.
322 See ioengine=net for more. If the ioengine is file based, you
323 can specify a number of files by separating the names with a
324 ':' colon. So if you wanted a job to open /dev/sda and /dev/sdb
325 as the two working files, you would use
326 filename=/dev/sda:/dev/sdb. On Windows, disk devices are
327 accessed as \\.\PhysicalDrive0 for the first device,
328 \\.\PhysicalDrive1 for the second etc. Note: Windows and
329 FreeBSD prevent write access to areas of the disk containing
330 in-use data (e.g. filesystems).
331 If the wanted filename does need to include a colon, then
332 escape that with a '\' character. For instance, if the filename
333 is "/dev/dsk/foo@3,0:c", then you would use
334 filename="/dev/dsk/foo@3,0\:c". '-' is a reserved name, meaning
335 stdin or stdout. Which of the two depends on the read/write
339 If sharing multiple files between jobs, it is usually necessary
340 to have fio generate the exact names that you want. By default,
341 fio will name a file based on the default file format
342 specification of jobname.jobnumber.filenumber. With this
343 option, that can be customized. Fio will recognize and replace
344 the following keywords in this string:
347 The name of the worker thread or process.
350 The incremental number of the worker thread or
354 The incremental number of the file for that worker
357 To have dependent jobs share a set of files, this option can
358 be set to have fio generate filenames that are shared between
359 the two. For instance, if testfiles.$filenum is specified,
360 file number 4 for any job will be named testfiles.4. The
361 default of $jobname.$jobnum.$filenum will be used if
362 no other format specifier is given.
364 opendir=str Tell fio to recursively add any file it can find in this
365 directory and down the file system tree.
367 lockfile=str Fio defaults to not locking any files before it does
368 IO to them. If a file or file descriptor is shared, fio
369 can serialize IO to that file to make the end result
370 consistent. This is usual for emulating real workloads that
371 share files. The lock modes are:
373 none No locking. The default.
374 exclusive Only one thread/process may do IO,
375 excluding all others.
376 readwrite Read-write locking on the file. Many
377 readers may access the file at the
378 same time, but writes get exclusive
382 rw=str Type of io pattern. Accepted values are:
384 read Sequential reads
385 write Sequential writes
386 randwrite Random writes
387 randread Random reads
388 rw,readwrite Sequential mixed reads and writes
389 randrw Random mixed reads and writes
391 For the mixed io types, the default is to split them 50/50.
392 For certain types of io the result may still be skewed a bit,
393 since the speed may be different. It is possible to specify
394 a number of IO's to do before getting a new offset, this is
395 done by appending a ':<nr>' to the end of the string given.
396 For a random read, it would look like 'rw=randread:8' for
397 passing in an offset modifier with a value of 8. If the
398 suffix is used with a sequential IO pattern, then the value
399 specified will be added to the generated offset for each IO.
400 For instance, using rw=write:4k will skip 4k for every
401 write. It turns sequential IO into sequential IO with holes.
402 See the 'rw_sequencer' option.
404 rw_sequencer=str If an offset modifier is given by appending a number to
405 the rw=<str> line, then this option controls how that
406 number modifies the IO offset being generated. Accepted
409 sequential Generate sequential offset
410 identical Generate the same offset
412 'sequential' is only useful for random IO, where fio would
413 normally generate a new random offset for every IO. If you
414 append eg 8 to randread, you would get a new random offset for
415 every 8 IO's. The result would be a seek for only every 8
416 IO's, instead of for every IO. Use rw=randread:8 to specify
417 that. As sequential IO is already sequential, setting
418 'sequential' for that would not result in any differences.
419 'identical' behaves in a similar fashion, except it sends
420 the same offset 8 number of times before generating a new
423 kb_base=int The base unit for a kilobyte. The defacto base is 2^10, 1024.
424 Storage manufacturers like to use 10^3 or 1000 as a base
425 ten unit instead, for obvious reasons. Allow values are
426 1024 or 1000, with 1024 being the default.
428 unified_rw_reporting=bool Fio normally reports statistics on a per
429 data direction basis, meaning that read, write, and trim are
430 accounted and reported separately. If this option is set,
431 the fio will sum the results and report them as "mixed"
434 randrepeat=bool For random IO workloads, seed the generator in a predictable
435 way so that results are repeatable across repetitions.
437 randseed=int Seed the random number generators based on this seed value, to
438 be able to control what sequence of output is being generated.
439 If not set, the random sequence depends on the randrepeat
442 fallocate=str Whether pre-allocation is performed when laying down files.
445 none Do not pre-allocate space
446 posix Pre-allocate via posix_fallocate()
447 keep Pre-allocate via fallocate() with
448 FALLOC_FL_KEEP_SIZE set
449 0 Backward-compatible alias for 'none'
450 1 Backward-compatible alias for 'posix'
452 May not be available on all supported platforms. 'keep' is only
453 available on Linux.If using ZFS on Solaris this must be set to
454 'none' because ZFS doesn't support it. Default: 'posix'.
456 fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
457 on what IO patterns it is likely to issue. Sometimes you
458 want to test specific IO patterns without telling the
459 kernel about it, in which case you can disable this option.
460 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
461 IO and POSIX_FADV_RANDOM for random IO.
463 size=int The total size of file io for this job. Fio will run until
464 this many bytes has been transferred, unless runtime is
465 limited by other options (such as 'runtime', for instance,
466 or increased/decreased by 'io_size'). Unless specific nrfiles
467 and filesize options are given, fio will divide this size
468 between the available files specified by the job. If not set,
469 fio will use the full size of the given files or devices.
470 If the files do not exist, size must be given. It is also
471 possible to give size as a percentage between 1 and 100. If
472 size=20% is given, fio will use 20% of the full size of the
473 given files or devices.
476 io_limit=int Normally fio operates within the region set by 'size', which
477 means that the 'size' option sets both the region and size of
478 IO to be performed. Sometimes that is not what you want. With
479 this option, it is possible to define just the amount of IO
480 that fio should do. For instance, if 'size' is set to 20G and
481 'io_size' is set to 5G, fio will perform IO within the first
482 20G but exit when 5G have been done. The opposite is also
483 possible - if 'size' is set to 20G, and 'io_size' is set to
484 40G, then fio will do 40G of IO within the 0..20G region.
486 filesize=int Individual file sizes. May be a range, in which case fio
487 will select sizes for files at random within the given range
488 and limited to 'size' in total (if that is given). If not
489 given, each created file is the same size.
491 file_append=bool Perform IO after the end of the file. Normally fio will
492 operate within the size of a file. If this option is set, then
493 fio will append to the file instead. This has identical
494 behavior to setting offset to the size of a file. This option
495 is ignored on non-regular files.
498 fill_fs=bool Sets size to something really large and waits for ENOSPC (no
499 space left on device) as the terminating condition. Only makes
500 sense with sequential write. For a read workload, the mount
501 point will be filled first then IO started on the result. This
502 option doesn't make sense if operating on a raw device node,
503 since the size of that is already known by the file system.
504 Additionally, writing beyond end-of-device will not return
508 bs=int The block size used for the io units. Defaults to 4k. Values
509 can be given for both read and writes. If a single int is
510 given, it will apply to both. If a second int is specified
511 after a comma, it will apply to writes only. In other words,
512 the format is either bs=read_and_write or bs=read,write,trim.
513 bs=4k,8k will thus use 4k blocks for reads, 8k blocks for
514 writes, and 8k for trims. You can terminate the list with
515 a trailing comma. bs=4k,8k, would use the default value for
516 trims.. If you only wish to set the write size, you
517 can do so by passing an empty read size - bs=,8k will set
518 8k for writes and leave the read default value.
521 ba=int At what boundary to align random IO offsets. Defaults to
522 the same as 'blocksize' the minimum blocksize given.
523 Minimum alignment is typically 512b for using direct IO,
524 though it usually depends on the hardware block size. This
525 option is mutually exclusive with using a random map for
526 files, so it will turn off that option.
528 blocksize_range=irange
529 bsrange=irange Instead of giving a single block size, specify a range
530 and fio will mix the issued io block sizes. The issued
531 io unit will always be a multiple of the minimum value
532 given (also see bs_unaligned). Applies to both reads and
533 writes, however a second range can be given after a comma.
536 bssplit=str Sometimes you want even finer grained control of the
537 block sizes issued, not just an even split between them.
538 This option allows you to weight various block sizes,
539 so that you are able to define a specific amount of
540 block sizes issued. The format for this option is:
542 bssplit=blocksize/percentage:blocksize/percentage
544 for as many block sizes as needed. So if you want to define
545 a workload that has 50% 64k blocks, 10% 4k blocks, and
546 40% 32k blocks, you would write:
548 bssplit=4k/10:64k/50:32k/40
550 Ordering does not matter. If the percentage is left blank,
551 fio will fill in the remaining values evenly. So a bssplit
552 option like this one:
554 bssplit=4k/50:1k/:32k/
556 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
557 always add up to 100, if bssplit is given a range that adds
558 up to more, it will error out.
560 bssplit also supports giving separate splits to reads and
561 writes. The format is identical to what bs= accepts. You
562 have to separate the read and write parts with a comma. So
563 if you want a workload that has 50% 2k reads and 50% 4k reads,
564 while having 90% 4k writes and 10% 8k writes, you would
567 bssplit=2k/50:4k/50,4k/90:8k/10
570 bs_unaligned If this option is given, any byte size value within bsrange
571 may be used as a block range. This typically wont work with
572 direct IO, as that normally requires sector alignment.
574 bs_is_seq_rand If this option is set, fio will use the normal read,write
575 blocksize settings as sequential,random instead. Any random
576 read or write will use the WRITE blocksize settings, and any
577 sequential read or write will use the READ blocksize setting.
579 zero_buffers If this option is given, fio will init the IO buffers to
580 all zeroes. The default is to fill them with random data.
581 The resulting IO buffers will not be completely zeroed,
582 unless scramble_buffers is also turned off.
584 refill_buffers If this option is given, fio will refill the IO buffers
585 on every submit. The default is to only fill it at init
586 time and reuse that data. Only makes sense if zero_buffers
587 isn't specified, naturally. If data verification is enabled,
588 refill_buffers is also automatically enabled.
590 scramble_buffers=bool If refill_buffers is too costly and the target is
591 using data deduplication, then setting this option will
592 slightly modify the IO buffer contents to defeat normal
593 de-dupe attempts. This is not enough to defeat more clever
594 block compression attempts, but it will stop naive dedupe of
595 blocks. Default: true.
597 buffer_compress_percentage=int If this is set, then fio will attempt to
598 provide IO buffer content (on WRITEs) that compress to
599 the specified level. Fio does this by providing a mix of
600 random data and a fixed pattern. The fixed pattern is either
601 zeroes, or the pattern specified by buffer_pattern. If the
602 pattern option is used, it might skew the compression ratio
603 slightly. Note that this is per block size unit, for file/disk
604 wide compression level that matches this setting, you'll also
605 want to set refill_buffers.
607 buffer_compress_chunk=int See buffer_compress_percentage. This
608 setting allows fio to manage how big the ranges of random
609 data and zeroed data is. Without this set, fio will
610 provide buffer_compress_percentage of blocksize random
611 data, followed by the remaining zeroed. With this set
612 to some chunk size smaller than the block size, fio can
613 alternate random and zeroed data throughout the IO
616 buffer_pattern=str If set, fio will fill the io buffers with this
617 pattern. If not set, the contents of io buffers is defined by
618 the other options related to buffer contents. The setting can
619 be any pattern of bytes, and can be prefixed with 0x for hex
620 values. It may also be a string, where the string must then
623 dedupe_percentage=int If set, fio will generate this percentage of
624 identical buffers when writing. These buffers will be
625 naturally dedupable. The contents of the buffers depend on
626 what other buffer compression settings have been set. It's
627 possible to have the individual buffers either fully
628 compressible, or not at all. This option only controls the
629 distribution of unique buffers.
631 nrfiles=int Number of files to use for this job. Defaults to 1.
633 openfiles=int Number of files to keep open at the same time. Defaults to
634 the same as nrfiles, can be set smaller to limit the number
637 file_service_type=str Defines how fio decides which file from a job to
638 service next. The following types are defined:
640 random Just choose a file at random.
642 roundrobin Round robin over open files. This
645 sequential Finish one file before moving on to
646 the next. Multiple files can still be
647 open depending on 'openfiles'.
649 The string can have a number appended, indicating how
650 often to switch to a new file. So if option random:4 is
651 given, fio will switch to a new random file after 4 ios
654 ioengine=str Defines how the job issues io to the file. The following
657 sync Basic read(2) or write(2) io. lseek(2) is
658 used to position the io location.
660 psync Basic pread(2) or pwrite(2) io.
662 vsync Basic readv(2) or writev(2) IO.
664 psyncv Basic preadv(2) or pwritev(2) IO.
666 libaio Linux native asynchronous io. Note that Linux
667 may only support queued behaviour with
668 non-buffered IO (set direct=1 or buffered=0).
669 This engine defines engine specific options.
671 posixaio glibc posix asynchronous io.
673 solarisaio Solaris native asynchronous io.
675 windowsaio Windows native asynchronous io.
677 mmap File is memory mapped and data copied
678 to/from using memcpy(3).
680 splice splice(2) is used to transfer the data and
681 vmsplice(2) to transfer data from user
684 syslet-rw Use the syslet system calls to make
685 regular read/write async.
687 sg SCSI generic sg v3 io. May either be
688 synchronous using the SG_IO ioctl, or if
689 the target is an sg character device
690 we use read(2) and write(2) for asynchronous
693 null Doesn't transfer any data, just pretends
694 to. This is mainly used to exercise fio
695 itself and for debugging/testing purposes.
697 net Transfer over the network to given host:port.
698 Depending on the protocol used, the hostname,
699 port, listen and filename options are used to
700 specify what sort of connection to make, while
701 the protocol option determines which protocol
703 This engine defines engine specific options.
705 netsplice Like net, but uses splice/vmsplice to
706 map data and send/receive.
707 This engine defines engine specific options.
709 cpuio Doesn't transfer any data, but burns CPU
710 cycles according to the cpuload= and
711 cpucycle= options. Setting cpuload=85
712 will cause that job to do nothing but burn
713 85% of the CPU. In case of SMP machines,
714 use numjobs=<no_of_cpu> to get desired CPU
715 usage, as the cpuload only loads a single
716 CPU at the desired rate.
718 guasi The GUASI IO engine is the Generic Userspace
719 Asyncronous Syscall Interface approach
722 http://www.xmailserver.org/guasi-lib.html
724 for more info on GUASI.
726 rdma The RDMA I/O engine supports both RDMA
727 memory semantics (RDMA_WRITE/RDMA_READ) and
728 channel semantics (Send/Recv) for the
729 InfiniBand, RoCE and iWARP protocols.
731 falloc IO engine that does regular fallocate to
732 simulate data transfer as fio ioengine.
733 DDIR_READ does fallocate(,mode = keep_size,)
734 DDIR_WRITE does fallocate(,mode = 0)
735 DDIR_TRIM does fallocate(,mode = punch_hole)
737 e4defrag IO engine that does regular EXT4_IOC_MOVE_EXT
738 ioctls to simulate defragment activity in
739 request to DDIR_WRITE event
741 rbd IO engine supporting direct access to Ceph
742 Rados Block Devices (RBD) via librbd without
743 the need to use the kernel rbd driver. This
744 ioengine defines engine specific options.
746 gfapi Using Glusterfs libgfapi sync interface to
747 direct access to Glusterfs volumes without
750 gfapi_async Using Glusterfs libgfapi async interface
751 to direct access to Glusterfs volumes without
752 having to go through FUSE. This ioengine
753 defines engine specific options.
755 libhdfs Read and write through Hadoop (HDFS).
756 The 'filename' option is used to specify host,
757 port of the hdfs name-node to connect. This
758 engine interprets offsets a little
759 differently. In HDFS, files once created
760 cannot be modified. So random writes are not
761 possible. To imitate this, libhdfs engine
762 expects bunch of small files to be created
763 over HDFS, and engine will randomly pick a
764 file out of those files based on the offset
765 generated by fio backend. (see the example
766 job file to create such files, use rw=write
767 option). Please note, you might want to set
768 necessary environment variables to work with
769 hdfs/libhdfs properly.
771 external Prefix to specify loading an external
772 IO engine object file. Append the engine
773 filename, eg ioengine=external:/tmp/foo.o
774 to load ioengine foo.o in /tmp.
776 iodepth=int This defines how many io units to keep in flight against
777 the file. The default is 1 for each file defined in this
778 job, can be overridden with a larger value for higher
779 concurrency. Note that increasing iodepth beyond 1 will not
780 affect synchronous ioengines (except for small degress when
781 verify_async is in use). Even async engines may impose OS
782 restrictions causing the desired depth not to be achieved.
783 This may happen on Linux when using libaio and not setting
784 direct=1, since buffered IO is not async on that OS. Keep an
785 eye on the IO depth distribution in the fio output to verify
786 that the achieved depth is as expected. Default: 1.
788 iodepth_batch_submit=int
789 iodepth_batch=int This defines how many pieces of IO to submit at once.
790 It defaults to 1 which means that we submit each IO
791 as soon as it is available, but can be raised to submit
792 bigger batches of IO at the time.
794 iodepth_batch_complete=int This defines how many pieces of IO to retrieve
795 at once. It defaults to 1 which means that we'll ask
796 for a minimum of 1 IO in the retrieval process from
797 the kernel. The IO retrieval will go on until we
798 hit the limit set by iodepth_low. If this variable is
799 set to 0, then fio will always check for completed
800 events before queuing more IO. This helps reduce
801 IO latency, at the cost of more retrieval system calls.
803 iodepth_low=int The low water mark indicating when to start filling
804 the queue again. Defaults to the same as iodepth, meaning
805 that fio will attempt to keep the queue full at all times.
806 If iodepth is set to eg 16 and iodepth_low is set to 4, then
807 after fio has filled the queue of 16 requests, it will let
808 the depth drain down to 4 before starting to fill it again.
810 direct=bool If value is true, use non-buffered io. This is usually
811 O_DIRECT. Note that ZFS on Solaris doesn't support direct io.
812 On Windows the synchronous ioengines don't support direct io.
814 atomic=bool If value is true, attempt to use atomic direct IO. Atomic
815 writes are guaranteed to be stable once acknowledged by
816 the operating system. Only Linux supports O_ATOMIC right
819 buffered=bool If value is true, use buffered io. This is the opposite
820 of the 'direct' option. Defaults to true.
822 offset=int Start io at the given offset in the file. The data before
823 the given offset will not be touched. This effectively
824 caps the file size at real_size - offset.
826 offset_increment=int If this is provided, then the real offset becomes
827 offset + offset_increment * thread_number, where the thread
828 number is a counter that starts at 0 and is incremented for
829 each sub-job (i.e. when numjobs option is specified). This
830 option is useful if there are several jobs which are intended
831 to operate on a file in parallel disjoint segments, with
832 even spacing between the starting points.
834 number_ios=int Fio will normally perform IOs until it has exhausted the size
835 of the region set by size=, or if it exhaust the allocated
836 time (or hits an error condition). With this setting, the
837 range/size can be set independently of the number of IOs to
838 perform. When fio reaches this number, it will exit normally
839 and report status. Note that this does not extend the amount
840 of IO that will be done, it will only stop fio if this
841 condition is met before other end-of-job criteria.
843 fsync=int If writing to a file, issue a sync of the dirty data
844 for every number of blocks given. For example, if you give
845 32 as a parameter, fio will sync the file for every 32
846 writes issued. If fio is using non-buffered io, we may
847 not sync the file. The exception is the sg io engine, which
848 synchronizes the disk cache anyway.
850 fdatasync=int Like fsync= but uses fdatasync() to only sync data and not
852 In FreeBSD and Windows there is no fdatasync(), this falls back to
855 sync_file_range=str:val Use sync_file_range() for every 'val' number of
856 write operations. Fio will track range of writes that
857 have happened since the last sync_file_range() call. 'str'
858 can currently be one or more of:
860 wait_before SYNC_FILE_RANGE_WAIT_BEFORE
861 write SYNC_FILE_RANGE_WRITE
862 wait_after SYNC_FILE_RANGE_WAIT_AFTER
864 So if you do sync_file_range=wait_before,write:8, fio would
865 use SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE for
866 every 8 writes. Also see the sync_file_range(2) man page.
867 This option is Linux specific.
869 overwrite=bool If true, writes to a file will always overwrite existing
870 data. If the file doesn't already exist, it will be
871 created before the write phase begins. If the file exists
872 and is large enough for the specified write phase, nothing
875 end_fsync=bool If true, fsync file contents when a write stage has completed.
877 fsync_on_close=bool If true, fio will fsync() a dirty file on close.
878 This differs from end_fsync in that it will happen on every
879 file close, not just at the end of the job.
881 rwmixread=int How large a percentage of the mix should be reads.
883 rwmixwrite=int How large a percentage of the mix should be writes. If both
884 rwmixread and rwmixwrite is given and the values do not add
885 up to 100%, the latter of the two will be used to override
886 the first. This may interfere with a given rate setting,
887 if fio is asked to limit reads or writes to a certain rate.
888 If that is the case, then the distribution may be skewed.
890 random_distribution=str:float By default, fio will use a completely uniform
891 random distribution when asked to perform random IO. Sometimes
892 it is useful to skew the distribution in specific ways,
893 ensuring that some parts of the data is more hot than others.
894 fio includes the following distribution models:
896 random Uniform random distribution
897 zipf Zipf distribution
898 pareto Pareto distribution
900 When using a zipf or pareto distribution, an input value
901 is also needed to define the access pattern. For zipf, this
902 is the zipf theta. For pareto, it's the pareto power. Fio
903 includes a test program, genzipf, that can be used visualize
904 what the given input values will yield in terms of hit rates.
905 If you wanted to use zipf with a theta of 1.2, you would use
906 random_distribution=zipf:1.2 as the option. If a non-uniform
907 model is used, fio will disable use of the random map.
909 percentage_random=int For a random workload, set how big a percentage should
910 be random. This defaults to 100%, in which case the workload
911 is fully random. It can be set from anywhere from 0 to 100.
912 Setting it to 0 would make the workload fully sequential. Any
913 setting in between will result in a random mix of sequential
914 and random IO, at the given percentages. It is possible to
915 set different values for reads, writes, and trim. To do so,
916 simply use a comma separated list. See blocksize.
918 norandommap Normally fio will cover every block of the file when doing
919 random IO. If this option is given, fio will just get a
920 new random offset without looking at past io history. This
921 means that some blocks may not be read or written, and that
922 some blocks may be read/written more than once. This option
923 is mutually exclusive with verify= if and only if multiple
924 blocksizes (via bsrange=) are used, since fio only tracks
925 complete rewrites of blocks.
927 softrandommap=bool See norandommap. If fio runs with the random block map
928 enabled and it fails to allocate the map, if this option is
929 set it will continue without a random block map. As coverage
930 will not be as complete as with random maps, this option is
933 random_generator=str Fio supports the following engines for generating
934 IO offsets for random IO:
936 tausworthe Strong 2^88 cycle random number generator
937 lfsr Linear feedback shift register generator
939 Tausworthe is a strong random number generator, but it
940 requires tracking on the side if we want to ensure that
941 blocks are only read or written once. LFSR guarantees
942 that we never generate the same offset twice, and it's
943 also less computationally expensive. It's not a true
944 random generator, however, though for IO purposes it's
945 typically good enough. LFSR only works with single
946 block sizes, not with workloads that use multiple block
947 sizes. If used with such a workload, fio may read or write
948 some blocks multiple times.
950 nice=int Run the job with the given nice value. See man nice(2).
952 prio=int Set the io priority value of this job. Linux limits us to
953 a positive value between 0 and 7, with 0 being the highest.
956 prioclass=int Set the io priority class. See man ionice(1).
958 thinktime=int Stall the job x microseconds after an io has completed before
959 issuing the next. May be used to simulate processing being
960 done by an application. See thinktime_blocks and
964 Only valid if thinktime is set - pretend to spend CPU time
965 doing something with the data received, before falling back
966 to sleeping for the rest of the period specified by
970 Only valid if thinktime is set - control how many blocks
971 to issue, before waiting 'thinktime' usecs. If not set,
972 defaults to 1 which will make fio wait 'thinktime' usecs
973 after every block. This effectively makes any queue depth
974 setting redundant, since no more than 1 IO will be queued
975 before we have to complete it and do our thinktime. In
976 other words, this setting effectively caps the queue depth
977 if the latter is larger.
979 rate=int Cap the bandwidth used by this job. The number is in bytes/sec,
980 the normal suffix rules apply. You can use rate=500k to limit
981 reads and writes to 500k each, or you can specify read and
982 writes separately. Using rate=1m,500k would limit reads to
983 1MB/sec and writes to 500KB/sec. Capping only reads or
984 writes can be done with rate=,500k or rate=500k,. The former
985 will only limit writes (to 500KB/sec), the latter will only
988 ratemin=int Tell fio to do whatever it can to maintain at least this
989 bandwidth. Failing to meet this requirement, will cause
990 the job to exit. The same format as rate is used for
991 read vs write separation.
993 rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
994 as rate, just specified independently of bandwidth. If the
995 job is given a block size range instead of a fixed value,
996 the smallest block size is used as the metric. The same format
997 as rate is used for read vs write separation.
999 rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
1000 the job to exit. The same format as rate is used for read vs
1003 latency_target=int If set, fio will attempt to find the max performance
1004 point that the given workload will run at while maintaining a
1005 latency below this target. The values is given in microseconds.
1006 See latency_window and latency_percentile
1008 latency_window=int Used with latency_target to specify the sample window
1009 that the job is run at varying queue depths to test the
1010 performance. The value is given in microseconds.
1012 latency_percentile=float The percentage of IOs that must fall within the
1013 criteria specified by latency_target and latency_window. If not
1014 set, this defaults to 100.0, meaning that all IOs must be equal
1015 or below to the value set by latency_target.
1017 max_latency=int If set, fio will exit the job if it exceeds this maximum
1018 latency. It will exit with an ETIME error.
1020 ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
1023 cpumask=int Set the CPU affinity of this job. The parameter given is a
1024 bitmask of allowed CPU's the job may run on. So if you want
1025 the allowed CPUs to be 1 and 5, you would pass the decimal
1026 value of (1 << 1 | 1 << 5), or 34. See man
1027 sched_setaffinity(2). This may not work on all supported
1028 operating systems or kernel versions. This option doesn't
1029 work well for a higher CPU count than what you can store in
1030 an integer mask, so it can only control cpus 1-32. For
1031 boxes with larger CPU counts, use cpus_allowed.
1033 cpus_allowed=str Controls the same options as cpumask, but it allows a text
1034 setting of the permitted CPUs instead. So to use CPUs 1 and
1035 5, you would specify cpus_allowed=1,5. This options also
1036 allows a range of CPUs. Say you wanted a binding to CPUs
1037 1, 5, and 8-15, you would set cpus_allowed=1,5,8-15.
1039 cpus_allowed_policy=str Set the policy of how fio distributes the CPUs
1040 specified by cpus_allowed or cpumask. Two policies are
1043 shared All jobs will share the CPU set specified.
1044 split Each job will get a unique CPU from the CPU set.
1046 'shared' is the default behaviour, if the option isn't
1047 specified. If split is specified, then fio will will assign
1048 one cpu per job. If not enough CPUs are given for the jobs
1049 listed, then fio will roundrobin the CPUs in the set.
1051 numa_cpu_nodes=str Set this job running on spcified NUMA nodes' CPUs. The
1052 arguments allow comma delimited list of cpu numbers,
1053 A-B ranges, or 'all'. Note, to enable numa options support,
1054 fio must be built on a system with libnuma-dev(el) installed.
1056 numa_mem_policy=str Set this job's memory policy and corresponding NUMA
1057 nodes. Format of the argements:
1059 `mode' is one of the following memory policy:
1060 default, prefer, bind, interleave, local
1061 For `default' and `local' memory policy, no node is
1062 needed to be specified.
1063 For `prefer', only one node is allowed.
1064 For `bind' and `interleave', it allow comma delimited
1065 list of numbers, A-B ranges, or 'all'.
1067 startdelay=time Start this job the specified number of seconds after fio
1068 has started. Only useful if the job file contains several
1069 jobs, and you want to delay starting some jobs to a certain
1072 runtime=time Tell fio to terminate processing after the specified number
1073 of seconds. It can be quite hard to determine for how long
1074 a specified job will run, so this parameter is handy to
1075 cap the total runtime to a given time.
1077 time_based If set, fio will run for the duration of the runtime
1078 specified even if the file(s) are completely read or
1079 written. It will simply loop over the same workload
1080 as many times as the runtime allows.
1082 ramp_time=time If set, fio will run the specified workload for this amount
1083 of time before logging any performance numbers. Useful for
1084 letting performance settle before logging results, thus
1085 minimizing the runtime required for stable results. Note
1086 that the ramp_time is considered lead in time for a job,
1087 thus it will increase the total runtime if a special timeout
1088 or runtime is specified.
1090 invalidate=bool Invalidate the buffer/page cache parts for this file prior
1091 to starting io. Defaults to true.
1093 sync=bool Use sync io for buffered writes. For the majority of the
1094 io engines, this means using O_SYNC.
1097 mem=str Fio can use various types of memory as the io unit buffer.
1098 The allowed values are:
1100 malloc Use memory from malloc(3) as the buffers.
1102 shm Use shared memory as the buffers. Allocated
1105 shmhuge Same as shm, but use huge pages as backing.
1107 mmap Use mmap to allocate buffers. May either be
1108 anonymous memory, or can be file backed if
1109 a filename is given after the option. The
1110 format is mem=mmap:/path/to/file.
1112 mmaphuge Use a memory mapped huge file as the buffer
1113 backing. Append filename after mmaphuge, ala
1114 mem=mmaphuge:/hugetlbfs/file
1116 The area allocated is a function of the maximum allowed
1117 bs size for the job, multiplied by the io depth given. Note
1118 that for shmhuge and mmaphuge to work, the system must have
1119 free huge pages allocated. This can normally be checked
1120 and set by reading/writing /proc/sys/vm/nr_hugepages on a
1121 Linux system. Fio assumes a huge page is 4MB in size. So
1122 to calculate the number of huge pages you need for a given
1123 job file, add up the io depth of all jobs (normally one unless
1124 iodepth= is used) and multiply by the maximum bs set. Then
1125 divide that number by the huge page size. You can see the
1126 size of the huge pages in /proc/meminfo. If no huge pages
1127 are allocated by having a non-zero number in nr_hugepages,
1128 using mmaphuge or shmhuge will fail. Also see hugepage-size.
1130 mmaphuge also needs to have hugetlbfs mounted and the file
1131 location should point there. So if it's mounted in /huge,
1132 you would use mem=mmaphuge:/huge/somefile.
1134 iomem_align=int This indiciates the memory alignment of the IO memory buffers.
1135 Note that the given alignment is applied to the first IO unit
1136 buffer, if using iodepth the alignment of the following buffers
1137 are given by the bs used. In other words, if using a bs that is
1138 a multiple of the page sized in the system, all buffers will
1139 be aligned to this value. If using a bs that is not page
1140 aligned, the alignment of subsequent IO memory buffers is the
1141 sum of the iomem_align and bs used.
1144 Defines the size of a huge page. Must at least be equal
1145 to the system setting, see /proc/meminfo. Defaults to 4MB.
1146 Should probably always be a multiple of megabytes, so using
1147 hugepage-size=Xm is the preferred way to set this to avoid
1148 setting a non-pow-2 bad value.
1150 exitall When one job finishes, terminate the rest. The default is
1151 to wait for each job to finish, sometimes that is not the
1154 bwavgtime=int Average the calculated bandwidth over the given time. Value
1155 is specified in milliseconds.
1157 iopsavgtime=int Average the calculated IOPS over the given time. Value
1158 is specified in milliseconds.
1160 create_serialize=bool If true, serialize the file creating for the jobs.
1161 This may be handy to avoid interleaving of data
1162 files, which may greatly depend on the filesystem
1163 used and even the number of processors in the system.
1165 create_fsync=bool fsync the data file after creation. This is the
1168 create_on_open=bool Don't pre-setup the files for IO, just create open()
1169 when it's time to do IO to that file.
1171 create_only=bool If true, fio will only run the setup phase of the job.
1172 If files need to be laid out or updated on disk, only
1173 that will be done. The actual job contents are not
1176 pre_read=bool If this is given, files will be pre-read into memory before
1177 starting the given IO operation. This will also clear
1178 the 'invalidate' flag, since it is pointless to pre-read
1179 and then drop the cache. This will only work for IO engines
1180 that are seekable, since they allow you to read the same data
1181 multiple times. Thus it will not work on eg network or splice
1184 unlink=bool Unlink the job files when done. Not the default, as repeated
1185 runs of that job would then waste time recreating the file
1186 set again and again.
1188 loops=int Run the specified number of iterations of this job. Used
1189 to repeat the same workload a given number of times. Defaults
1192 verify_only Do not perform specified workload---only verify data still
1193 matches previous invocation of this workload. This option
1194 allows one to check data multiple times at a later date
1195 without overwriting it. This option makes sense only for
1196 workloads that write data, and does not support workloads
1197 with the time_based option set.
1199 do_verify=bool Run the verify phase after a write phase. Only makes sense if
1200 verify is set. Defaults to 1.
1202 verify=str If writing to a file, fio can verify the file contents
1203 after each iteration of the job. The allowed values are:
1205 md5 Use an md5 sum of the data area and store
1206 it in the header of each block.
1208 crc64 Use an experimental crc64 sum of the data
1209 area and store it in the header of each
1212 crc32c Use a crc32c sum of the data area and store
1213 it in the header of each block.
1215 crc32c-intel Use hardware assisted crc32c calcuation
1216 provided on SSE4.2 enabled processors. Falls
1217 back to regular software crc32c, if not
1218 supported by the system.
1220 crc32 Use a crc32 sum of the data area and store
1221 it in the header of each block.
1223 crc16 Use a crc16 sum of the data area and store
1224 it in the header of each block.
1226 crc7 Use a crc7 sum of the data area and store
1227 it in the header of each block.
1229 xxhash Use xxhash as the checksum function. Generally
1230 the fastest software checksum that fio
1233 sha512 Use sha512 as the checksum function.
1235 sha256 Use sha256 as the checksum function.
1237 sha1 Use optimized sha1 as the checksum function.
1239 meta Write extra information about each io
1240 (timestamp, block number etc.). The block
1241 number is verified. The io sequence number is
1242 verified for workloads that write data.
1243 See also verify_pattern.
1245 null Only pretend to verify. Useful for testing
1246 internals with ioengine=null, not for much
1249 This option can be used for repeated burn-in tests of a
1250 system to make sure that the written data is also
1251 correctly read back. If the data direction given is
1252 a read or random read, fio will assume that it should
1253 verify a previously written file. If the data direction
1254 includes any form of write, the verify will be of the
1257 verifysort=bool If set, fio will sort written verify blocks when it deems
1258 it faster to read them back in a sorted manner. This is
1259 often the case when overwriting an existing file, since
1260 the blocks are already laid out in the file system. You
1261 can ignore this option unless doing huge amounts of really
1262 fast IO where the red-black tree sorting CPU time becomes
1265 verify_offset=int Swap the verification header with data somewhere else
1266 in the block before writing. Its swapped back before
1269 verify_interval=int Write the verification header at a finer granularity
1270 than the blocksize. It will be written for chunks the
1271 size of header_interval. blocksize should divide this
1274 verify_pattern=str If set, fio will fill the io buffers with this
1275 pattern. Fio defaults to filling with totally random
1276 bytes, but sometimes it's interesting to fill with a known
1277 pattern for io verification purposes. Depending on the
1278 width of the pattern, fio will fill 1/2/3/4 bytes of the
1279 buffer at the time(it can be either a decimal or a hex number).
1280 The verify_pattern if larger than a 32-bit quantity has to
1281 be a hex number that starts with either "0x" or "0X". Use
1284 verify_fatal=bool Normally fio will keep checking the entire contents
1285 before quitting on a block verification failure. If this
1286 option is set, fio will exit the job on the first observed
1289 verify_dump=bool If set, dump the contents of both the original data
1290 block and the data block we read off disk to files. This
1291 allows later analysis to inspect just what kind of data
1292 corruption occurred. Off by default.
1294 verify_async=int Fio will normally verify IO inline from the submitting
1295 thread. This option takes an integer describing how many
1296 async offload threads to create for IO verification instead,
1297 causing fio to offload the duty of verifying IO contents
1298 to one or more separate threads. If using this offload
1299 option, even sync IO engines can benefit from using an
1300 iodepth setting higher than 1, as it allows them to have
1301 IO in flight while verifies are running.
1303 verify_async_cpus=str Tell fio to set the given CPU affinity on the
1304 async IO verification threads. See cpus_allowed for the
1307 verify_backlog=int Fio will normally verify the written contents of a
1308 job that utilizes verify once that job has completed. In
1309 other words, everything is written then everything is read
1310 back and verified. You may want to verify continually
1311 instead for a variety of reasons. Fio stores the meta data
1312 associated with an IO block in memory, so for large
1313 verify workloads, quite a bit of memory would be used up
1314 holding this meta data. If this option is enabled, fio
1315 will write only N blocks before verifying these blocks.
1317 verify_backlog_batch=int Control how many blocks fio will verify
1318 if verify_backlog is set. If not set, will default to
1319 the value of verify_backlog (meaning the entire queue
1320 is read back and verified). If verify_backlog_batch is
1321 less than verify_backlog then not all blocks will be verified,
1322 if verify_backlog_batch is larger than verify_backlog, some
1323 blocks will be verified more than once.
1325 verify_state_save=bool When a job exits during the write phase of a verify
1326 workload, save its current state. This allows fio to replay
1327 up until that point, if the verify state is loaded for the
1328 verify read phase. The format of the filename is, roughly,
1329 <type>-<jobname>-<jobindex>-verify.state. <type> is "local"
1330 for a local run, "sock" for a client/server socket connection,
1331 and "ip" (192.168.0.1, for instance) for a networked
1332 client/server connection.
1334 verify_state_load=bool If a verify termination trigger was used, fio stores
1335 the current write state of each thread. This can be used at
1336 verification time so that fio knows how far it should verify.
1337 Without this information, fio will run a full verification
1338 pass, according to the settings in the job file used.
1341 wait_for_previous Wait for preceding jobs in the job file to exit, before
1342 starting this one. Can be used to insert serialization
1343 points in the job file. A stone wall also implies starting
1344 a new reporting group.
1346 new_group Start a new reporting group. See: group_reporting.
1348 numjobs=int Create the specified number of clones of this job. May be
1349 used to setup a larger number of threads/processes doing
1350 the same thing. Each thread is reported separately; to see
1351 statistics for all clones as a whole, use group_reporting in
1352 conjunction with new_group.
1354 group_reporting It may sometimes be interesting to display statistics for
1355 groups of jobs as a whole instead of for each individual job.
1356 This is especially true if 'numjobs' is used; looking at
1357 individual thread/process output quickly becomes unwieldy.
1358 To see the final report per-group instead of per-job, use
1359 'group_reporting'. Jobs in a file will be part of the same
1360 reporting group, unless if separated by a stonewall, or by
1363 thread fio defaults to forking jobs, however if this option is
1364 given, fio will use pthread_create(3) to create threads
1367 zonesize=int Divide a file into zones of the specified size. See zoneskip.
1369 zoneskip=int Skip the specified number of bytes when zonesize data has
1370 been read. The two zone options can be used to only do
1371 io on zones of a file.
1373 write_iolog=str Write the issued io patterns to the specified file. See
1374 read_iolog. Specify a separate file for each job, otherwise
1375 the iologs will be interspersed and the file may be corrupt.
1377 read_iolog=str Open an iolog with the specified file name and replay the
1378 io patterns it contains. This can be used to store a
1379 workload and replay it sometime later. The iolog given
1380 may also be a blktrace binary file, which allows fio
1381 to replay a workload captured by blktrace. See blktrace
1382 for how to capture such logging data. For blktrace replay,
1383 the file needs to be turned into a blkparse binary data
1384 file first (blkparse <device> -o /dev/null -d file_for_fio.bin).
1386 replay_no_stall=int When replaying I/O with read_iolog the default behavior
1387 is to attempt to respect the time stamps within the log and
1388 replay them with the appropriate delay between IOPS. By
1389 setting this variable fio will not respect the timestamps and
1390 attempt to replay them as fast as possible while still
1391 respecting ordering. The result is the same I/O pattern to a
1392 given device, but different timings.
1394 replay_redirect=str While replaying I/O patterns using read_iolog the
1395 default behavior is to replay the IOPS onto the major/minor
1396 device that each IOP was recorded from. This is sometimes
1397 undesirable because on a different machine those major/minor
1398 numbers can map to a different device. Changing hardware on
1399 the same system can also result in a different major/minor
1400 mapping. Replay_redirect causes all IOPS to be replayed onto
1401 the single specified device regardless of the device it was
1402 recorded from. i.e. replay_redirect=/dev/sdc would cause all
1403 IO in the blktrace to be replayed onto /dev/sdc. This means
1404 multiple devices will be replayed onto a single, if the trace
1405 contains multiple devices. If you want multiple devices to be
1406 replayed concurrently to multiple redirected devices you must
1407 blkparse your trace into separate traces and replay them with
1408 independent fio invocations. Unfortuantely this also breaks
1409 the strict time ordering between multiple device accesses.
1411 write_bw_log=str If given, write a bandwidth log of the jobs in this job
1412 file. Can be used to store data of the bandwidth of the
1413 jobs in their lifetime. The included fio_generate_plots
1414 script uses gnuplot to turn these text files into nice
1415 graphs. See write_lat_log for behaviour of given
1416 filename. For this option, the suffix is _bw.x.log, where
1417 x is the index of the job (1..N, where N is the number of
1420 write_lat_log=str Same as write_bw_log, except that this option stores io
1421 submission, completion, and total latencies instead. If no
1422 filename is given with this option, the default filename of
1423 "jobname_type.log" is used. Even if the filename is given,
1424 fio will still append the type of log. So if one specifies
1428 The actual log names will be foo_slat.x.log, foo_clat.x.log,
1429 and foo_lat.x.log, where x is the index of the job (1..N,
1430 where N is the number of jobs). This helps fio_generate_plot
1431 fine the logs automatically.
1433 write_iops_log=str Same as write_bw_log, but writes IOPS. If no filename is
1434 given with this option, the default filename of
1435 "jobname_type.x.log" is used,where x is the index of the job
1436 (1..N, where N is the number of jobs). Even if the filename
1437 is given, fio will still append the type of log.
1439 log_avg_msec=int By default, fio will log an entry in the iops, latency,
1440 or bw log for every IO that completes. When writing to the
1441 disk log, that can quickly grow to a very large size. Setting
1442 this option makes fio average the each log entry over the
1443 specified period of time, reducing the resolution of the log.
1446 log_offset=int If this is set, the iolog options will include the byte
1447 offset for the IO entry as well as the other data values.
1449 log_compression=int If this is set, fio will compress the IO logs as
1450 it goes, to keep the memory footprint lower. When a log
1451 reaches the specified size, that chunk is removed and
1452 compressed in the background. Given that IO logs are
1453 fairly highly compressible, this yields a nice memory
1454 savings for longer runs. The downside is that the
1455 compression will consume some background CPU cycles, so
1456 it may impact the run. This, however, is also true if
1457 the logging ends up consuming most of the system memory.
1458 So pick your poison. The IO logs are saved normally at the
1459 end of a run, by decompressing the chunks and storing them
1460 in the specified log file. This feature depends on the
1461 availability of zlib.
1463 log_store_compressed=bool If set, and log_compression is also set,
1464 fio will store the log files in a compressed format. They
1465 can be decompressed with fio, using the --inflate-log
1466 command line parameter. The files will be stored with a
1469 lockmem=int Pin down the specified amount of memory with mlock(2). Can
1470 potentially be used instead of removing memory or booting
1471 with less memory to simulate a smaller amount of memory.
1472 The amount specified is per worker.
1474 exec_prerun=str Before running this job, issue the command specified
1475 through system(3). Output is redirected in a file called
1478 exec_postrun=str After the job completes, issue the command specified
1479 though system(3). Output is redirected in a file called
1480 jobname.postrun.txt.
1482 ioscheduler=str Attempt to switch the device hosting the file to the specified
1483 io scheduler before running.
1485 disk_util=bool Generate disk utilization statistics, if the platform
1486 supports it. Defaults to on.
1488 disable_lat=bool Disable measurements of total latency numbers. Useful
1489 only for cutting back the number of calls to gettimeofday,
1490 as that does impact performance at really high IOPS rates.
1491 Note that to really get rid of a large amount of these
1492 calls, this option must be used with disable_slat and
1495 disable_clat=bool Disable measurements of completion latency numbers. See
1498 disable_slat=bool Disable measurements of submission latency numbers. See
1501 disable_bw=bool Disable measurements of throughput/bandwidth numbers. See
1504 clat_percentiles=bool Enable the reporting of percentiles of
1505 completion latencies.
1507 percentile_list=float_list Overwrite the default list of percentiles
1508 for completion latencies. Each number is a floating
1509 number in the range (0,100], and the maximum length of
1510 the list is 20. Use ':' to separate the numbers, and
1511 list the numbers in ascending order. For example,
1512 --percentile_list=99.5:99.9 will cause fio to report
1513 the values of completion latency below which 99.5% and
1514 99.9% of the observed latencies fell, respectively.
1516 clocksource=str Use the given clocksource as the base of timing. The
1517 supported options are:
1519 gettimeofday gettimeofday(2)
1521 clock_gettime clock_gettime(2)
1523 cpu Internal CPU clock source
1525 cpu is the preferred clocksource if it is reliable, as it
1526 is very fast (and fio is heavy on time calls). Fio will
1527 automatically use this clocksource if it's supported and
1528 considered reliable on the system it is running on, unless
1529 another clocksource is specifically set. For x86/x86-64 CPUs,
1530 this means supporting TSC Invariant.
1532 gtod_reduce=bool Enable all of the gettimeofday() reducing options
1533 (disable_clat, disable_slat, disable_bw) plus reduce
1534 precision of the timeout somewhat to really shrink
1535 the gettimeofday() call count. With this option enabled,
1536 we only do about 0.4% of the gtod() calls we would have
1537 done if all time keeping was enabled.
1539 gtod_cpu=int Sometimes it's cheaper to dedicate a single thread of
1540 execution to just getting the current time. Fio (and
1541 databases, for instance) are very intensive on gettimeofday()
1542 calls. With this option, you can set one CPU aside for
1543 doing nothing but logging current time to a shared memory
1544 location. Then the other threads/processes that run IO
1545 workloads need only copy that segment, instead of entering
1546 the kernel with a gettimeofday() call. The CPU set aside
1547 for doing these time calls will be excluded from other
1548 uses. Fio will manually clear it from the CPU mask of other
1551 continue_on_error=str Normally fio will exit the job on the first observed
1552 failure. If this option is set, fio will continue the job when
1553 there is a 'non-fatal error' (EIO or EILSEQ) until the runtime
1554 is exceeded or the I/O size specified is completed. If this
1555 option is used, there are two more stats that are appended,
1556 the total error count and the first error. The error field
1557 given in the stats is the first error that was hit during the
1560 The allowed values are:
1562 none Exit on any IO or verify errors.
1564 read Continue on read errors, exit on all others.
1566 write Continue on write errors, exit on all others.
1568 io Continue on any IO error, exit on all others.
1570 verify Continue on verify errors, exit on all others.
1572 all Continue on all errors.
1574 0 Backward-compatible alias for 'none'.
1576 1 Backward-compatible alias for 'all'.
1578 ignore_error=str Sometimes you want to ignore some errors during test
1579 in that case you can specify error list for each error type.
1580 ignore_error=READ_ERR_LIST,WRITE_ERR_LIST,VERIFY_ERR_LIST
1581 errors for given error type is separated with ':'. Error
1582 may be symbol ('ENOSPC', 'ENOMEM') or integer.
1584 ignore_error=EAGAIN,ENOSPC:122
1585 This option will ignore EAGAIN from READ, and ENOSPC and
1586 122(EDQUOT) from WRITE.
1588 error_dump=bool If set dump every error even if it is non fatal, true
1589 by default. If disabled only fatal error will be dumped
1591 cgroup=str Add job to this control group. If it doesn't exist, it will
1592 be created. The system must have a mounted cgroup blkio
1593 mount point for this to work. If your system doesn't have it
1594 mounted, you can do so with:
1596 # mount -t cgroup -o blkio none /cgroup
1598 cgroup_weight=int Set the weight of the cgroup to this value. See
1599 the documentation that comes with the kernel, allowed values
1600 are in the range of 100..1000.
1602 cgroup_nodelete=bool Normally fio will delete the cgroups it has created after
1603 the job completion. To override this behavior and to leave
1604 cgroups around after the job completion, set cgroup_nodelete=1.
1605 This can be useful if one wants to inspect various cgroup
1606 files after job completion. Default: false
1608 uid=int Instead of running as the invoking user, set the user ID to
1609 this value before the thread/process does any work.
1611 gid=int Set group ID, see uid.
1613 flow_id=int The ID of the flow. If not specified, it defaults to being a
1614 global flow. See flow.
1616 flow=int Weight in token-based flow control. If this value is used, then
1617 there is a 'flow counter' which is used to regulate the
1618 proportion of activity between two or more jobs. fio attempts
1619 to keep this flow counter near zero. The 'flow' parameter
1620 stands for how much should be added or subtracted to the flow
1621 counter on each iteration of the main I/O loop. That is, if
1622 one job has flow=8 and another job has flow=-1, then there
1623 will be a roughly 1:8 ratio in how much one runs vs the other.
1625 flow_watermark=int The maximum value that the absolute value of the flow
1626 counter is allowed to reach before the job must wait for a
1627 lower value of the counter.
1629 flow_sleep=int The period of time, in microseconds, to wait after the flow
1630 watermark has been exceeded before retrying operations
1632 In addition, there are some parameters which are only valid when a specific
1633 ioengine is in use. These are used identically to normal parameters, with the
1634 caveat that when used on the command line, they must come after the ioengine
1635 that defines them is selected.
1637 [libaio] userspace_reap Normally, with the libaio engine in use, fio will use
1638 the io_getevents system call to reap newly returned events.
1639 With this flag turned on, the AIO ring will be read directly
1640 from user-space to reap events. The reaping mode is only
1641 enabled when polling for a minimum of 0 events (eg when
1642 iodepth_batch_complete=0).
1644 [cpu] cpuload=int Attempt to use the specified percentage of CPU cycles.
1646 [cpu] cpuchunks=int Split the load into cycles of the given time. In
1649 [cpu] exit_on_io_done=bool Detect when IO threads are done, then exit.
1651 [netsplice] hostname=str
1652 [net] hostname=str The host name or IP address to use for TCP or UDP based IO.
1653 If the job is a TCP listener or UDP reader, the hostname is not
1654 used and must be omitted unless it is a valid UDP multicast
1657 [netsplice] port=int
1658 [net] port=int The TCP or UDP port to bind to or connect to. If this is used
1659 with numjobs to spawn multiple instances of the same job type, then this will
1660 be the starting port number since fio will use a range of ports.
1662 [netsplice] interface=str
1663 [net] interface=str The IP address of the network interface used to send or
1664 receive UDP multicast
1667 [net] ttl=int Time-to-live value for outgoing UDP multicast packets.
1670 [netsplice] nodelay=bool
1671 [net] nodelay=bool Set TCP_NODELAY on TCP connections.
1673 [netsplice] protocol=str
1674 [netsplice] proto=str
1676 [net] proto=str The network protocol to use. Accepted values are:
1678 tcp Transmission control protocol
1679 tcpv6 Transmission control protocol V6
1680 udp User datagram protocol
1681 udpv6 User datagram protocol V6
1682 unix UNIX domain socket
1684 When the protocol is TCP or UDP, the port must also be given,
1685 as well as the hostname if the job is a TCP listener or UDP
1686 reader. For unix sockets, the normal filename option should be
1687 used and the port is invalid.
1689 [net] listen For TCP network connections, tell fio to listen for incoming
1690 connections rather than initiating an outgoing connection. The
1691 hostname must be omitted if this option is used.
1693 [net] pingpong Normaly a network writer will just continue writing data, and
1694 a network reader will just consume packages. If pingpong=1
1695 is set, a writer will send its normal payload to the reader,
1696 then wait for the reader to send the same payload back. This
1697 allows fio to measure network latencies. The submission
1698 and completion latencies then measure local time spent
1699 sending or receiving, and the completion latency measures
1700 how long it took for the other end to receive and send back.
1701 For UDP multicast traffic pingpong=1 should only be set for a
1702 single reader when multiple readers are listening to the same
1705 [net] window_size Set the desired socket buffer size for the connection.
1707 [net] mss Set the TCP maximum segment size (TCP_MAXSEG).
1709 [e4defrag] donorname=str
1710 File will be used as a block donor(swap extents between files)
1711 [e4defrag] inplace=int
1712 Configure donor file blocks allocation strategy
1713 0(default): Preallocate donor's file on init
1714 1 : allocate space immidietly inside defragment event,
1715 and free right after event
1719 6.0 Interpreting the output
1720 ---------------------------
1722 fio spits out a lot of output. While running, fio will display the
1723 status of the jobs created. An example of that would be:
1725 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1727 The characters inside the square brackets denote the current status of
1728 each thread. The possible values (in typical life cycle order) are:
1732 P Thread setup, but not started.
1734 I Thread initialized, waiting or generating necessary data.
1735 p Thread running pre-reading file(s).
1736 R Running, doing sequential reads.
1737 r Running, doing random reads.
1738 W Running, doing sequential writes.
1739 w Running, doing random writes.
1740 M Running, doing mixed sequential reads/writes.
1741 m Running, doing mixed random reads/writes.
1742 F Running, currently waiting for fsync()
1743 f Running, finishing up (writing IO logs, etc)
1744 V Running, doing verification of written data.
1745 E Thread exited, not reaped by main thread yet.
1747 X Thread reaped, exited with an error.
1748 K Thread reaped, exited due to signal.
1750 Fio will condense the thread string as not to take up more space on the
1751 command line as is needed. For instance, if you have 10 readers and 10
1752 writers running, the output would look like this:
1754 Jobs: 20 (f=20): [R(10),W(10)] [4.0% done] [2103MB/0KB/0KB /s] [538K/0/0 iops] [eta 57m:36s]
1756 Fio will still maintain the ordering, though. So the above means that jobs
1757 1..10 are readers, and 11..20 are writers.
1759 The other values are fairly self explanatory - number of threads
1760 currently running and doing io, rate of io since last check (read speed
1761 listed first, then write speed), and the estimated completion percentage
1762 and time for the running group. It's impossible to estimate runtime of
1763 the following groups (if any). Note that the string is displayed in order,
1764 so it's possible to tell which of the jobs are currently doing what. The
1765 first character is the first job defined in the job file, and so forth.
1767 When fio is done (or interrupted by ctrl-c), it will show the data for
1768 each thread, group of threads, and disks in that order. For each data
1769 direction, the output looks like:
1771 Client1 (g=0): err= 0:
1772 write: io= 32MB, bw= 666KB/s, iops=89 , runt= 50320msec
1773 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
1774 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
1775 bw (KB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
1776 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
1777 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
1778 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1779 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1780 issued r/w: total=0/32768, short=0/0
1781 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
1782 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
1784 The client number is printed, along with the group id and error of that
1785 thread. Below is the io statistics, here for writes. In the order listed,
1788 io= Number of megabytes io performed
1789 bw= Average bandwidth rate
1790 iops= Average IOs performed per second
1791 runt= The runtime of that thread
1792 slat= Submission latency (avg being the average, stdev being the
1793 standard deviation). This is the time it took to submit
1794 the io. For sync io, the slat is really the completion
1795 latency, since queue/complete is one operation there. This
1796 value can be in milliseconds or microseconds, fio will choose
1797 the most appropriate base and print that. In the example
1798 above, milliseconds is the best scale. Note: in --minimal mode
1799 latencies are always expressed in microseconds.
1800 clat= Completion latency. Same names as slat, this denotes the
1801 time from submission to completion of the io pieces. For
1802 sync io, clat will usually be equal (or very close) to 0,
1803 as the time from submit to complete is basically just
1804 CPU time (io has already been done, see slat explanation).
1805 bw= Bandwidth. Same names as the xlat stats, but also includes
1806 an approximate percentage of total aggregate bandwidth
1807 this thread received in this group. This last value is
1808 only really useful if the threads in this group are on the
1809 same disk, since they are then competing for disk access.
1810 cpu= CPU usage. User and system time, along with the number
1811 of context switches this thread went through, usage of
1812 system and user time, and finally the number of major
1813 and minor page faults.
1814 IO depths= The distribution of io depths over the job life time. The
1815 numbers are divided into powers of 2, so for example the
1816 16= entries includes depths up to that value but higher
1817 than the previous entry. In other words, it covers the
1818 range from 16 to 31.
1819 IO submit= How many pieces of IO were submitting in a single submit
1820 call. Each entry denotes that amount and below, until
1821 the previous entry - eg, 8=100% mean that we submitted
1822 anywhere in between 5-8 ios per submit call.
1823 IO complete= Like the above submit number, but for completions instead.
1824 IO issued= The number of read/write requests issued, and how many
1826 IO latencies= The distribution of IO completion latencies. This is the
1827 time from when IO leaves fio and when it gets completed.
1828 The numbers follow the same pattern as the IO depths,
1829 meaning that 2=1.6% means that 1.6% of the IO completed
1830 within 2 msecs, 20=12.8% means that 12.8% of the IO
1831 took more than 10 msecs, but less than (or equal to) 20 msecs.
1833 After each client has been listed, the group statistics are printed. They
1834 will look like this:
1836 Run status group 0 (all jobs):
1837 READ: io=64MB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
1838 WRITE: io=64MB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
1840 For each data direction, it prints:
1842 io= Number of megabytes io performed.
1843 aggrb= Aggregate bandwidth of threads in this group.
1844 minb= The minimum average bandwidth a thread saw.
1845 maxb= The maximum average bandwidth a thread saw.
1846 mint= The smallest runtime of the threads in that group.
1847 maxt= The longest runtime of the threads in that group.
1849 And finally, the disk statistics are printed. They will look like this:
1851 Disk stats (read/write):
1852 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
1854 Each value is printed for both reads and writes, with reads first. The
1857 ios= Number of ios performed by all groups.
1858 merge= Number of merges io the io scheduler.
1859 ticks= Number of ticks we kept the disk busy.
1860 io_queue= Total time spent in the disk queue.
1861 util= The disk utilization. A value of 100% means we kept the disk
1862 busy constantly, 50% would be a disk idling half of the time.
1864 It is also possible to get fio to dump the current output while it is
1865 running, without terminating the job. To do that, send fio the USR1 signal.
1866 You can also get regularly timed dumps by using the --status-interval
1867 parameter, or by creating a file in /tmp named fio-dump-status. If fio
1868 sees this file, it will unlink it and dump the current output status.
1874 For scripted usage where you typically want to generate tables or graphs
1875 of the results, fio can output the results in a semicolon separated format.
1876 The format is one long line of values, such as:
1878 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%
1879 A description of this job goes here.
1881 The job description (if provided) follows on a second line.
1883 To enable terse output, use the --minimal command line option. The first
1884 value is the version of the terse output format. If the output has to
1885 be changed for some reason, this number will be incremented by 1 to
1886 signify that change.
1888 Split up, the format is as follows:
1890 terse version, fio version, jobname, groupid, error
1892 Total IO (KB), bandwidth (KB/sec), IOPS, runtime (msec)
1893 Submission latency: min, max, mean, deviation (usec)
1894 Completion latency: min, max, mean, deviation (usec)
1895 Completion latency percentiles: 20 fields (see below)
1896 Total latency: min, max, mean, deviation (usec)
1897 Bw (KB/s): min, max, aggregate percentage of total, mean, deviation
1899 Total IO (KB), bandwidth (KB/sec), IOPS, runtime (msec)
1900 Submission latency: min, max, mean, deviation (usec)
1901 Completion latency: min, max, mean, deviation (usec)
1902 Completion latency percentiles: 20 fields (see below)
1903 Total latency: min, max, mean, deviation (usec)
1904 Bw (KB/s): min, max, aggregate percentage of total, mean, deviation
1905 CPU usage: user, system, context switches, major faults, minor faults
1906 IO depths: <=1, 2, 4, 8, 16, 32, >=64
1907 IO latencies microseconds: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1908 IO latencies milliseconds: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1909 Disk utilization: Disk name, Read ios, write ios,
1910 Read merges, write merges,
1911 Read ticks, write ticks,
1912 Time spent in queue, disk utilization percentage
1913 Additional Info (dependent on continue_on_error, default off): total # errors, first error code
1915 Additional Info (dependent on description being set): Text description
1917 Completion latency percentiles can be a grouping of up to 20 sets, so
1918 for the terse output fio writes all of them. Each field will look like this:
1922 which is the Xth percentile, and the usec latency associated with it.
1924 For disk utilization, all disks used by fio are shown. So for each disk
1925 there will be a disk utilization section.
1928 8.0 Trace file format
1929 ---------------------
1930 There are two trace file format that you can encounter. The older (v1) format
1931 is unsupported since version 1.20-rc3 (March 2008). It will still be described
1932 below in case that you get an old trace and want to understand it.
1934 In any case the trace is a simple text file with a single action per line.
1937 8.1 Trace file format v1
1938 ------------------------
1939 Each line represents a single io action in the following format:
1943 where rw=0/1 for read/write, and the offset and length entries being in bytes.
1945 This format is not supported in Fio versions => 1.20-rc3.
1948 8.2 Trace file format v2
1949 ------------------------
1950 The second version of the trace file format was added in Fio version 1.17.
1951 It allows to access more then one file per trace and has a bigger set of
1952 possible file actions.
1954 The first line of the trace file has to be:
1958 Following this can be lines in two different formats, which are described below.
1960 The file management format:
1964 The filename is given as an absolute path. The action can be one of these:
1966 add Add the given filename to the trace
1967 open Open the file with the given filename. The filename has to have
1968 been added with the add action before.
1969 close Close the file with the given filename. The file has to have been
1973 The file io action format:
1975 filename action offset length
1977 The filename is given as an absolute path, and has to have been added and opened
1978 before it can be used with this format. The offset and length are given in
1979 bytes. The action can be one of these:
1981 wait Wait for 'offset' microseconds. Everything below 100 is discarded.
1982 read Read 'length' bytes beginning from 'offset'
1983 write Write 'length' bytes beginning from 'offset'
1984 sync fsync() the file
1985 datasync fdatasync() the file
1986 trim trim the given file from the given 'offset' for 'length' bytes
1989 9.0 CPU idleness profiling
1990 --------------------------
1991 In some cases, we want to understand CPU overhead in a test. For example,
1992 we test patches for the specific goodness of whether they reduce CPU usage.
1993 fio implements a balloon approach to create a thread per CPU that runs at
1994 idle priority, meaning that it only runs when nobody else needs the cpu.
1995 By measuring the amount of work completed by the thread, idleness of each
1996 CPU can be derived accordingly.
1998 An unit work is defined as touching a full page of unsigned characters. Mean
1999 and standard deviation of time to complete an unit work is reported in "unit
2000 work" section. Options can be chosen to report detailed percpu idleness or
2001 overall system idleness by aggregating percpu stats.
2004 10.0 Verification and triggers
2005 ------------------------------
2006 Fio is usually run in one of two ways, when data verification is done. The
2007 first is a normal write job of some sort with verify enabled. When the
2008 write phase has completed, fio switches to reads and verifies everything
2009 it wrote. The second model is running just the write phase, and then later
2010 on running the same job (but with reads instead of writes) to repeat the
2011 same IO patterns and verify the contents. Both of these methods depend
2012 on the write phase being completed, as fio otherwise has no idea how much
2015 With verification triggers, fio supports dumping the current write state
2016 to local files. Then a subsequent read verify workload can load this state
2017 and know exactly where to stop. This is useful for testing cases where
2018 power is cut to a server in a managed fashion, for instance.
2020 A verification trigger consists of two things:
2022 1) Storing the write state of each job
2023 2) Executing a trigger command
2025 The write state is relatively small, on the order of hundreds of bytes
2026 to single kilobytes. It contains information on the number of completions
2027 done, the last X completions, etc.
2029 A trigger is invoked either through creation ('touch') of a specified
2030 file in the system, or through a timeout setting. If fio is run with
2031 --trigger-file=/tmp/trigger-file, then it will continually check for
2032 the existence of /tmp/trigger-file. When it sees this file, it will
2033 fire off the trigger (thus saving state, and executing the trigger
2036 For client/server runs, there's both a local and remote trigger. If
2037 fio is running as a server backend, it will send the job states back
2038 to the client for safe storage, then execute the remote trigger, if
2039 specified. If a local trigger is specified, the server will still send
2040 back the write state, but the client will then execute the trigger.
2042 10.1 Verification trigger example
2043 ---------------------------------
2044 Lets say we want to run a powercut test on the remote machine 'server'.
2045 Our write workload is in write-test.fio. We want to cut power to 'server'
2046 at some point during the run, and we'll run this test from the safety
2047 or our local machine, 'localbox'. On the server, we'll start the fio
2050 server# fio --server
2052 and on the client, we'll fire off the workload:
2054 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger-remote="bash -c \"echo b > /proc/sysrq-triger\""
2056 We set /tmp/my-trigger as the trigger file, and we tell fio to execute
2058 echo b > /proc/sysrq-trigger
2060 on the server once it has received the trigger and sent us the write
2061 state. This will work, but it's not _really_ cutting power to the server,
2062 it's merely abruptly rebooting it. If we have a remote way of cutting
2063 power to the server through IPMI or similar, we could do that through
2064 a local trigger command instead. Lets assume we have a script that does
2065 IPMI reboot of a given hostname, ipmi-reboot. On localbox, we could
2066 then have run fio with a local trigger instead:
2068 localbox$ fio --client=server --trigger-file=/tmp/my-trigger --trigger="ipmi-reboot server"
2070 For this case, fio would wait for the server to send us the write state,
2071 then execute 'ipmi-reboot server' when that happened.
2073 10.1 Loading verify state
2074 -------------------------
2075 To load store write state, read verification job file must contain
2076 the verify_state_load option. If that is set, fio will load the previously
2077 stored state. For a local fio run this is done by loading the files directly,
2078 and on a client/server run, the server backend will ask the client to send
2079 the files over and load them from there.