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 fio also supports environment variable expansion in job files. Any
162 substring of the form "${VARNAME}" as part of an option value (in other
163 words, on the right of the `='), will be expanded to the value of the
164 environment variable called VARNAME. If no such environment variable
165 is defined, or VARNAME is the empty string, the empty string will be
168 As an example, let's look at a sample fio invocation and job file:
170 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
172 ; -- start job file --
179 This will expand to the following equivalent job file at runtime:
181 ; -- start job file --
188 fio ships with a few example job files, you can also look there for
192 5.0 Detailed list of parameters
193 -------------------------------
195 This section describes in details each parameter associated with a job.
196 Some parameters take an option of a given type, such as an integer or
197 a string. The following types are used:
199 str String. This is a sequence of alpha characters.
200 time Integer with possible time suffix. In seconds unless otherwise
201 specified, use eg 10m for 10 minutes. Accepts s/m/h for seconds,
203 int SI integer. A whole number value, which may contain a suffix
204 describing the base of the number. Accepted suffixes are k/m/g/t/p,
205 meaning kilo, mega, giga, tera, and peta. The suffix is not case
206 sensitive. So if you want to specify 4096, you could either write
207 out '4096' or just give 4k. The suffixes signify base 2 values, so
208 1024 is 1k and 1024k is 1m and so on. If the option accepts an upper
209 and lower range, use a colon ':' or minus '-' to separate such values.
210 May also include a prefix to indicate numbers base. If 0x is used,
211 the number is assumed to be hexadecimal. See irange.
212 bool Boolean. Usually parsed as an integer, however only defined for
213 true and false (1 and 0).
214 irange Integer range with suffix. Allows value range to be given, such
215 as 1024-4096. A colon may also be used as the separator, eg
216 1k:4k. If the option allows two sets of ranges, they can be
217 specified with a ',' or '/' delimiter: 1k-4k/8k-32k. Also see
220 With the above in mind, here follows the complete list of fio job
223 name=str ASCII name of the job. This may be used to override the
224 name printed by fio for this job. Otherwise the job
225 name is used. On the command line this parameter has the
226 special purpose of also signaling the start of a new
229 description=str Text description of the job. Doesn't do anything except
230 dump this text description when this job is run. It's
233 directory=str Prefix filenames with this directory. Used to place files
234 in a different location than "./".
236 filename=str Fio normally makes up a filename based on the job name,
237 thread number, and file number. If you want to share
238 files between threads in a job or several jobs, specify
239 a filename for each of them to override the default. If
240 the ioengine used is 'net', the filename is the host, port,
241 and protocol to use in the format of =host/port/protocol.
242 See ioengine=net for more. If the ioengine is file based, you
243 can specify a number of files by separating the names with a
244 ':' colon. So if you wanted a job to open /dev/sda and /dev/sdb
245 as the two working files, you would use
246 filename=/dev/sda:/dev/sdb. '-' is a reserved name, meaning
247 stdin or stdout. Which of the two depends on the read/write
250 opendir=str Tell fio to recursively add any file it can find in this
251 directory and down the file system tree.
253 lockfile=str Fio defaults to not locking any files before it does
254 IO to them. If a file or file descriptor is shared, fio
255 can serialize IO to that file to make the end result
256 consistent. This is usual for emulating real workloads that
257 share files. The lock modes are:
259 none No locking. The default.
260 exclusive Only one thread/process may do IO,
261 excluding all others.
262 readwrite Read-write locking on the file. Many
263 readers may access the file at the
264 same time, but writes get exclusive
267 The option may be post-fixed with a lock batch number. If
268 set, then each thread/process may do that amount of IOs to
269 the file before giving up the lock. Since lock acquisition is
270 expensive, batching the lock/unlocks will speed up IO.
273 rw=str Type of io pattern. Accepted values are:
275 read Sequential reads
276 write Sequential writes
277 randwrite Random writes
278 randread Random reads
279 rw Sequential mixed reads and writes
280 randrw Random mixed reads and writes
282 For the mixed io types, the default is to split them 50/50.
283 For certain types of io the result may still be skewed a bit,
284 since the speed may be different. It is possible to specify
285 a number of IO's to do before getting a new offset - this
286 is only useful for random IO, where fio would normally
287 generate a new random offset for every IO. If you append
288 eg 8 to randread, you would get a new random offset for
289 every 8 IO's. The result would be a seek for only every 8
290 IO's, instead of for every IO. Use rw=randread:8 to specify
293 kb_base=int The base unit for a kilobyte. The defacto base is 2^10, 1024.
294 Storage manufacturers like to use 10^3 or 1000 as a base
295 ten unit instead, for obvious reasons. Allow values are
296 1024 or 1000, with 1024 being the default.
298 randrepeat=bool For random IO workloads, seed the generator in a predictable
299 way so that results are repeatable across repetitions.
301 fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
302 on what IO patterns it is likely to issue. Sometimes you
303 want to test specific IO patterns without telling the
304 kernel about it, in which case you can disable this option.
305 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
306 IO and POSIX_FADV_RANDOM for random IO.
308 size=int The total size of file io for this job. Fio will run until
309 this many bytes has been transferred, unless runtime is
310 limited by other options (such as 'runtime', for instance).
311 Unless specific nrfiles and filesize options are given,
312 fio will divide this size between the available files
313 specified by the job.
315 filesize=int Individual file sizes. May be a range, in which case fio
316 will select sizes for files at random within the given range
317 and limited to 'size' in total (if that is given). If not
318 given, each created file is the same size.
320 fill_device=bool Sets size to something really large and waits for ENOSPC (no
321 space left on device) as the terminating condition. Only makes
322 sense with sequential write. For a read workload, the mount
323 point will be filled first then IO started on the result.
326 bs=int The block size used for the io units. Defaults to 4k. Values
327 can be given for both read and writes. If a single int is
328 given, it will apply to both. If a second int is specified
329 after a comma, it will apply to writes only. In other words,
330 the format is either bs=read_and_write or bs=read,write.
331 bs=4k,8k will thus use 4k blocks for reads, and 8k blocks
332 for writes. If you only wish to set the write size, you
333 can do so by passing an empty read size - bs=,8k will set
334 8k for writes and leave the read default value.
337 ba=int At what boundary to align random IO offsets. Defaults to
338 the same as 'blocksize' the minimum blocksize given.
339 Minimum alignment is typically 512b for using direct IO,
340 though it usually depends on the hardware block size. This
341 option is mutually exclusive with using a random map for
342 files, so it will turn off that option.
344 blocksize_range=irange
345 bsrange=irange Instead of giving a single block size, specify a range
346 and fio will mix the issued io block sizes. The issued
347 io unit will always be a multiple of the minimum value
348 given (also see bs_unaligned). Applies to both reads and
349 writes, however a second range can be given after a comma.
352 bssplit=str Sometimes you want even finer grained control of the
353 block sizes issued, not just an even split between them.
354 This option allows you to weight various block sizes,
355 so that you are able to define a specific amount of
356 block sizes issued. The format for this option is:
358 bssplit=blocksize/percentage:blocksize/percentage
360 for as many block sizes as needed. So if you want to define
361 a workload that has 50% 64k blocks, 10% 4k blocks, and
362 40% 32k blocks, you would write:
364 bssplit=4k/10:64k/50:32k/40
366 Ordering does not matter. If the percentage is left blank,
367 fio will fill in the remaining values evenly. So a bssplit
368 option like this one:
370 bssplit=4k/50:1k/:32k/
372 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
373 always add up to 100, if bssplit is given a range that adds
374 up to more, it will error out.
376 bssplit also supports giving separate splits to reads and
377 writes. The format is identical to what bs= accepts. You
378 have to separate the read and write parts with a comma. So
379 if you want a workload that has 50% 2k reads and 50% 4k reads,
380 while having 90% 4k writes and 10% 8k writes, you would
383 bssplit=2k/50:4k/50,4k/90,8k/10
386 bs_unaligned If this option is given, any byte size value within bsrange
387 may be used as a block range. This typically wont work with
388 direct IO, as that normally requires sector alignment.
390 zero_buffers If this option is given, fio will init the IO buffers to
391 all zeroes. The default is to fill them with random data.
393 refill_buffers If this option is given, fio will refill the IO buffers
394 on every submit. The default is to only fill it at init
395 time and reuse that data. Only makes sense if zero_buffers
396 isn't specified, naturally. If data verification is enabled,
397 refill_buffers is also automatically enabled.
399 nrfiles=int Number of files to use for this job. Defaults to 1.
401 openfiles=int Number of files to keep open at the same time. Defaults to
402 the same as nrfiles, can be set smaller to limit the number
405 file_service_type=str Defines how fio decides which file from a job to
406 service next. The following types are defined:
408 random Just choose a file at random.
410 roundrobin Round robin over open files. This
413 sequential Finish one file before moving on to
414 the next. Multiple files can still be
415 open depending on 'openfiles'.
417 The string can have a number appended, indicating how
418 often to switch to a new file. So if option random:4 is
419 given, fio will switch to a new random file after 4 ios
422 ioengine=str Defines how the job issues io to the file. The following
425 sync Basic read(2) or write(2) io. lseek(2) is
426 used to position the io location.
428 psync Basic pread(2) or pwrite(2) io.
430 vsync Basic readv(2) or writev(2) IO.
432 libaio Linux native asynchronous io. Note that Linux
433 may only support queued behaviour with
434 non-buffered IO (set direct=1 or buffered=0).
436 posixaio glibc posix asynchronous io.
438 solarisaio Solaris native asynchronous io.
440 mmap File is memory mapped and data copied
441 to/from using memcpy(3).
443 splice splice(2) is used to transfer the data and
444 vmsplice(2) to transfer data from user
447 syslet-rw Use the syslet system calls to make
448 regular read/write async.
450 sg SCSI generic sg v3 io. May either be
451 synchronous using the SG_IO ioctl, or if
452 the target is an sg character device
453 we use read(2) and write(2) for asynchronous
456 null Doesn't transfer any data, just pretends
457 to. This is mainly used to exercise fio
458 itself and for debugging/testing purposes.
460 net Transfer over the network to given host:port.
461 'filename' must be set appropriately to
462 filename=host/port/protocol regardless of send
463 or receive, if the latter only the port
464 argument is used. 'host' may be an IP address
465 or hostname, port is the port number to be used,
466 and protocol may be 'udp' or 'tcp'. If no
467 protocol is given, TCP is used.
469 netsplice Like net, but uses splice/vmsplice to
470 map data and send/receive.
472 cpuio Doesn't transfer any data, but burns CPU
473 cycles according to the cpuload= and
474 cpucycle= options. Setting cpuload=85
475 will cause that job to do nothing but burn
476 85% of the CPU. In case of SMP machines,
477 use numjobs=<no_of_cpu> to get desired CPU
478 usage, as the cpuload only loads a single
479 CPU at the desired rate.
481 guasi The GUASI IO engine is the Generic Userspace
482 Asyncronous Syscall Interface approach
485 http://www.xmailserver.org/guasi-lib.html
487 for more info on GUASI.
489 external Prefix to specify loading an external
490 IO engine object file. Append the engine
491 filename, eg ioengine=external:/tmp/foo.o
492 to load ioengine foo.o in /tmp.
494 iodepth=int This defines how many io units to keep in flight against
495 the file. The default is 1 for each file defined in this
496 job, can be overridden with a larger value for higher
499 iodepth_batch_submit=int
500 iodepth_batch=int This defines how many pieces of IO to submit at once.
501 It defaults to 1 which means that we submit each IO
502 as soon as it is available, but can be raised to submit
503 bigger batches of IO at the time.
505 iodepth_batch_complete=int This defines how many pieces of IO to retrieve
506 at once. It defaults to 1 which means that we'll ask
507 for a minimum of 1 IO in the retrieval process from
508 the kernel. The IO retrieval will go on until we
509 hit the limit set by iodepth_low. If this variable is
510 set to 0, then fio will always check for completed
511 events before queuing more IO. This helps reduce
512 IO latency, at the cost of more retrieval system calls.
514 iodepth_low=int The low water mark indicating when to start filling
515 the queue again. Defaults to the same as iodepth, meaning
516 that fio will attempt to keep the queue full at all times.
517 If iodepth is set to eg 16 and iodepth_low is set to 4, then
518 after fio has filled the queue of 16 requests, it will let
519 the depth drain down to 4 before starting to fill it again.
521 direct=bool If value is true, use non-buffered io. This is usually
524 buffered=bool If value is true, use buffered io. This is the opposite
525 of the 'direct' option. Defaults to true.
527 offset=int Start io at the given offset in the file. The data before
528 the given offset will not be touched. This effectively
529 caps the file size at real_size - offset.
531 fsync=int If writing to a file, issue a sync of the dirty data
532 for every number of blocks given. For example, if you give
533 32 as a parameter, fio will sync the file for every 32
534 writes issued. If fio is using non-buffered io, we may
535 not sync the file. The exception is the sg io engine, which
536 synchronizes the disk cache anyway.
538 fsyncdata=int Like fsync= but uses fdatasync() to only sync data and not
541 overwrite=bool If true, writes to a file will always overwrite existing
542 data. If the file doesn't already exist, it will be
543 created before the write phase begins. If the file exists
544 and is large enough for the specified write phase, nothing
547 end_fsync=bool If true, fsync file contents when the job exits.
549 fsync_on_close=bool If true, fio will fsync() a dirty file on close.
550 This differs from end_fsync in that it will happen on every
551 file close, not just at the end of the job.
553 rwmixread=int How large a percentage of the mix should be reads.
555 rwmixwrite=int How large a percentage of the mix should be writes. If both
556 rwmixread and rwmixwrite is given and the values do not add
557 up to 100%, the latter of the two will be used to override
558 the first. This may interfere with a given rate setting,
559 if fio is asked to limit reads or writes to a certain rate.
560 If that is the case, then the distribution may be skewed.
562 norandommap Normally fio will cover every block of the file when doing
563 random IO. If this option is given, fio will just get a
564 new random offset without looking at past io history. This
565 means that some blocks may not be read or written, and that
566 some blocks may be read/written more than once. This option
567 is mutually exclusive with verify= if and only if multiple
568 blocksizes (via bsrange=) are used, since fio only tracks
569 complete rewrites of blocks.
571 softrandommap See norandommap. If fio runs with the random block map enabled
572 and it fails to allocate the map, if this option is set it
573 will continue without a random block map. As coverage will
574 not be as complete as with random maps, this option is
577 nice=int Run the job with the given nice value. See man nice(2).
579 prio=int Set the io priority value of this job. Linux limits us to
580 a positive value between 0 and 7, with 0 being the highest.
583 prioclass=int Set the io priority class. See man ionice(1).
585 thinktime=int Stall the job x microseconds after an io has completed before
586 issuing the next. May be used to simulate processing being
587 done by an application. See thinktime_blocks and
591 Only valid if thinktime is set - pretend to spend CPU time
592 doing something with the data received, before falling back
593 to sleeping for the rest of the period specified by
597 Only valid if thinktime is set - control how many blocks
598 to issue, before waiting 'thinktime' usecs. If not set,
599 defaults to 1 which will make fio wait 'thinktime' usecs
602 rate=int Cap the bandwidth used by this job. The number is in bytes/sec,
603 the normal suffix rules apply. You can use rate=500k to limit
604 reads and writes to 500k each, or you can specify read and
605 writes separately. Using rate=1m,500k would limit reads to
606 1MB/sec and writes to 500KB/sec. Capping only reads or
607 writes can be done with rate=,500k or rate=500k,. The former
608 will only limit writes (to 500KB/sec), the latter will only
611 ratemin=int Tell fio to do whatever it can to maintain at least this
612 bandwidth. Failing to meet this requirement, will cause
613 the job to exit. The same format as rate is used for
614 read vs write separation.
616 rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
617 as rate, just specified independently of bandwidth. If the
618 job is given a block size range instead of a fixed value,
619 the smallest block size is used as the metric. The same format
620 as rate is used for read vs write seperation.
622 rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
623 the job to exit. The same format as rate is used for read vs
626 ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
629 cpumask=int Set the CPU affinity of this job. The parameter given is a
630 bitmask of allowed CPU's the job may run on. So if you want
631 the allowed CPUs to be 1 and 5, you would pass the decimal
632 value of (1 << 1 | 1 << 5), or 34. See man
633 sched_setaffinity(2). This may not work on all supported
634 operating systems or kernel versions. This option doesn't
635 work well for a higher CPU count than what you can store in
636 an integer mask, so it can only control cpus 1-32. For
637 boxes with larger CPU counts, use cpus_allowed.
639 cpus_allowed=str Controls the same options as cpumask, but it allows a text
640 setting of the permitted CPUs instead. So to use CPUs 1 and
641 5, you would specify cpus_allowed=1,5. This options also
642 allows a range of CPUs. Say you wanted a binding to CPUs
643 1, 5, and 8-15, you would set cpus_allowed=1,5,8-15.
645 startdelay=time Start this job the specified number of seconds after fio
646 has started. Only useful if the job file contains several
647 jobs, and you want to delay starting some jobs to a certain
650 runtime=time Tell fio to terminate processing after the specified number
651 of seconds. It can be quite hard to determine for how long
652 a specified job will run, so this parameter is handy to
653 cap the total runtime to a given time.
655 time_based If set, fio will run for the duration of the runtime
656 specified even if the file(s) are completely read or
657 written. It will simply loop over the same workload
658 as many times as the runtime allows.
660 ramp_time=time If set, fio will run the specified workload for this amount
661 of time before logging any performance numbers. Useful for
662 letting performance settle before logging results, thus
663 minimizing the runtime required for stable results. Note
664 that the ramp_time is considered lead in time for a job,
665 thus it will increase the total runtime if a special timeout
666 or runtime is specified.
668 invalidate=bool Invalidate the buffer/page cache parts for this file prior
669 to starting io. Defaults to true.
671 sync=bool Use sync io for buffered writes. For the majority of the
672 io engines, this means using O_SYNC.
675 mem=str Fio can use various types of memory as the io unit buffer.
676 The allowed values are:
678 malloc Use memory from malloc(3) as the buffers.
680 shm Use shared memory as the buffers. Allocated
683 shmhuge Same as shm, but use huge pages as backing.
685 mmap Use mmap to allocate buffers. May either be
686 anonymous memory, or can be file backed if
687 a filename is given after the option. The
688 format is mem=mmap:/path/to/file.
690 mmaphuge Use a memory mapped huge file as the buffer
691 backing. Append filename after mmaphuge, ala
692 mem=mmaphuge:/hugetlbfs/file
694 The area allocated is a function of the maximum allowed
695 bs size for the job, multiplied by the io depth given. Note
696 that for shmhuge and mmaphuge to work, the system must have
697 free huge pages allocated. This can normally be checked
698 and set by reading/writing /proc/sys/vm/nr_hugepages on a
699 Linux system. Fio assumes a huge page is 4MB in size. So
700 to calculate the number of huge pages you need for a given
701 job file, add up the io depth of all jobs (normally one unless
702 iodepth= is used) and multiply by the maximum bs set. Then
703 divide that number by the huge page size. You can see the
704 size of the huge pages in /proc/meminfo. If no huge pages
705 are allocated by having a non-zero number in nr_hugepages,
706 using mmaphuge or shmhuge will fail. Also see hugepage-size.
708 mmaphuge also needs to have hugetlbfs mounted and the file
709 location should point there. So if it's mounted in /huge,
710 you would use mem=mmaphuge:/huge/somefile.
712 iomem_align=int This indiciates the memory alignment of the IO memory buffers.
713 Note that the given alignment is applied to the first IO unit
714 buffer, if using iodepth the alignment of the following buffers
715 are given by the bs used. In other words, if using a bs that is
716 a multiple of the page sized in the system, all buffers will
717 be aligned to this value. If using a bs that is not page
718 aligned, the alignment of subsequent IO memory buffers is the
719 sum of the iomem_align and bs used.
722 Defines the size of a huge page. Must at least be equal
723 to the system setting, see /proc/meminfo. Defaults to 4MB.
724 Should probably always be a multiple of megabytes, so using
725 hugepage-size=Xm is the preferred way to set this to avoid
726 setting a non-pow-2 bad value.
728 exitall When one job finishes, terminate the rest. The default is
729 to wait for each job to finish, sometimes that is not the
732 bwavgtime=int Average the calculated bandwidth over the given time. Value
733 is specified in milliseconds.
735 create_serialize=bool If true, serialize the file creating for the jobs.
736 This may be handy to avoid interleaving of data
737 files, which may greatly depend on the filesystem
738 used and even the number of processors in the system.
740 create_fsync=bool fsync the data file after creation. This is the
743 create_on_open=bool Don't pre-setup the files for IO, just create open()
744 when it's time to do IO to that file.
746 pre_read=bool If this is given, files will be pre-read into memory before
747 starting the given IO operation. This will also clear
748 the 'invalidate' flag, since it is pointless to pre-read
749 and then drop the cache. This will only work for IO engines
750 that are seekable, since they allow you to read the same data
751 multiple times. Thus it will not work on eg network or splice
754 unlink=bool Unlink the job files when done. Not the default, as repeated
755 runs of that job would then waste time recreating the file
758 loops=int Run the specified number of iterations of this job. Used
759 to repeat the same workload a given number of times. Defaults
762 do_verify=bool Run the verify phase after a write phase. Only makes sense if
763 verify is set. Defaults to 1.
765 verify=str If writing to a file, fio can verify the file contents
766 after each iteration of the job. The allowed values are:
768 md5 Use an md5 sum of the data area and store
769 it in the header of each block.
771 crc64 Use an experimental crc64 sum of the data
772 area and store it in the header of each
775 crc32c Use a crc32c sum of the data area and store
776 it in the header of each block.
778 crc32c-intel Use hardware assisted crc32c calcuation
779 provided on SSE4.2 enabled processors.
781 crc32 Use a crc32 sum of the data area and store
782 it in the header of each block.
784 crc16 Use a crc16 sum of the data area and store
785 it in the header of each block.
787 crc7 Use a crc7 sum of the data area and store
788 it in the header of each block.
790 sha512 Use sha512 as the checksum function.
792 sha256 Use sha256 as the checksum function.
794 meta Write extra information about each io
795 (timestamp, block number etc.). The block
798 null Only pretend to verify. Useful for testing
799 internals with ioengine=null, not for much
802 This option can be used for repeated burn-in tests of a
803 system to make sure that the written data is also
806 verifysort=bool If set, fio will sort written verify blocks when it deems
807 it faster to read them back in a sorted manner. This is
808 often the case when overwriting an existing file, since
809 the blocks are already laid out in the file system. You
810 can ignore this option unless doing huge amounts of really
811 fast IO where the red-black tree sorting CPU time becomes
814 verify_offset=int Swap the verification header with data somewhere else
815 in the block before writing. Its swapped back before
818 verify_interval=int Write the verification header at a finer granularity
819 than the blocksize. It will be written for chunks the
820 size of header_interval. blocksize should divide this
823 verify_pattern=int If set, fio will fill the io buffers with this
824 pattern. Fio defaults to filling with totally random
825 bytes, but sometimes it's interesting to fill with a known
826 pattern for io verification purposes. Depending on the
827 width of the pattern, fio will fill 1/2/3/4 bytes of the
828 buffer at the time. The verify_pattern cannot be larger than
831 verify_fatal=bool Normally fio will keep checking the entire contents
832 before quitting on a block verification failure. If this
833 option is set, fio will exit the job on the first observed
836 verify_async=int Fio will normally verify IO inline from the submitting
837 thread. This option takes an integer describing how many
838 async offload threads to create for IO verification instead,
839 causing fio to offload the duty of verifying IO contents
840 to one or more separate threads. If using this offload
841 option, even sync IO engines can benefit from using an
842 iodepth setting higher than 1, as it allows them to have
843 IO in flight while verifies are running.
845 verify_async_cpus=str Tell fio to set the given CPU affinity on the
846 async IO verification threads. See cpus_allowed for the
849 stonewall Wait for preceeding jobs in the job file to exit, before
850 starting this one. Can be used to insert serialization
851 points in the job file. A stone wall also implies starting
852 a new reporting group.
854 new_group Start a new reporting group. If this option isn't given,
855 jobs in a file will be part of the same reporting group
856 unless separated by a stone wall (or if it's a group
857 by itself, with the numjobs option).
859 numjobs=int Create the specified number of clones of this job. May be
860 used to setup a larger number of threads/processes doing
861 the same thing. We regard that grouping of jobs as a
864 group_reporting If 'numjobs' is set, it may be interesting to display
865 statistics for the group as a whole instead of for each
866 individual job. This is especially true of 'numjobs' is
867 large, looking at individual thread/process output quickly
868 becomes unwieldy. If 'group_reporting' is specified, fio
869 will show the final report per-group instead of per-job.
871 thread fio defaults to forking jobs, however if this option is
872 given, fio will use pthread_create(3) to create threads
875 zonesize=int Divide a file into zones of the specified size. See zoneskip.
877 zoneskip=int Skip the specified number of bytes when zonesize data has
878 been read. The two zone options can be used to only do
879 io on zones of a file.
881 write_iolog=str Write the issued io patterns to the specified file. See
884 read_iolog=str Open an iolog with the specified file name and replay the
885 io patterns it contains. This can be used to store a
886 workload and replay it sometime later. The iolog given
887 may also be a blktrace binary file, which allows fio
888 to replay a workload captured by blktrace. See blktrace
889 for how to capture such logging data. For blktrace replay,
890 the file needs to be turned into a blkparse binary data
891 file first (blktrace <device> -d file_for_fio.bin).
893 write_bw_log=str If given, write a bandwidth log of the jobs in this job
894 file. Can be used to store data of the bandwidth of the
895 jobs in their lifetime. The included fio_generate_plots
896 script uses gnuplot to turn these text files into nice
897 graphs. See write_log_log for behaviour of given
898 filename. For this option, the postfix is _bw.log.
900 write_lat_log=str Same as write_bw_log, except that this option stores io
901 completion latencies instead. If no filename is given
902 with this option, the default filename of "jobname_type.log"
903 is used. Even if the filename is given, fio will still
904 append the type of log. So if one specifies
908 The actual log names will be foo_clat.log and foo_slat.log.
909 This helps fio_generate_plot fine the logs automatically.
911 lockmem=int Pin down the specified amount of memory with mlock(2). Can
912 potentially be used instead of removing memory or booting
913 with less memory to simulate a smaller amount of memory.
915 exec_prerun=str Before running this job, issue the command specified
918 exec_postrun=str After the job completes, issue the command specified
921 ioscheduler=str Attempt to switch the device hosting the file to the specified
922 io scheduler before running.
924 cpuload=int If the job is a CPU cycle eater, attempt to use the specified
925 percentage of CPU cycles.
927 cpuchunks=int If the job is a CPU cycle eater, split the load into
928 cycles of the given time. In microseconds.
930 disk_util=bool Generate disk utilization statistics, if the platform
931 supports it. Defaults to on.
933 disable_clat=bool Disable measurements of completion latency numbers. Useful
934 only for cutting back the number of calls to gettimeofday,
935 as that does impact performance at really high IOPS rates.
936 Note that to really get rid of a large amount of these
937 calls, this option must be used with disable_slat and
940 disable_slat=bool Disable measurements of submission latency numbers. See
943 disable_bw=bool Disable measurements of throughput/bandwidth numbers. See
946 gtod_reduce=bool Enable all of the gettimeofday() reducing options
947 (disable_clat, disable_slat, disable_bw) plus reduce
948 precision of the timeout somewhat to really shrink
949 the gettimeofday() call count. With this option enabled,
950 we only do about 0.4% of the gtod() calls we would have
951 done if all time keeping was enabled.
953 gtod_cpu=int Sometimes it's cheaper to dedicate a single thread of
954 execution to just getting the current time. Fio (and
955 databases, for instance) are very intensive on gettimeofday()
956 calls. With this option, you can set one CPU aside for
957 doing nothing but logging current time to a shared memory
958 location. Then the other threads/processes that run IO
959 workloads need only copy that segment, instead of entering
960 the kernel with a gettimeofday() call. The CPU set aside
961 for doing these time calls will be excluded from other
962 uses. Fio will manually clear it from the CPU mask of other
964 continue_on_error=bool Normally fio will exit the job on the first observed
965 failure. If this option is set, fio will continue the job when
966 there is a 'non-fatal error' (EIO or EILSEQ) until the runtime
967 is exceeded or the I/O size specified is completed. If this
968 option is used, there are two more stats that are appended,
969 the total error count and the first error. The error field
970 given in the stats is the first error that was hit during the
974 6.0 Interpreting the output
975 ---------------------------
977 fio spits out a lot of output. While running, fio will display the
978 status of the jobs created. An example of that would be:
980 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
982 The characters inside the square brackets denote the current status of
983 each thread. The possible values (in typical life cycle order) are:
987 P Thread setup, but not started.
989 I Thread initialized, waiting.
990 p Thread running pre-reading file(s).
991 R Running, doing sequential reads.
992 r Running, doing random reads.
993 W Running, doing sequential writes.
994 w Running, doing random writes.
995 M Running, doing mixed sequential reads/writes.
996 m Running, doing mixed random reads/writes.
997 F Running, currently waiting for fsync()
998 V Running, doing verification of written data.
999 E Thread exited, not reaped by main thread yet.
1002 The other values are fairly self explanatory - number of threads
1003 currently running and doing io, rate of io since last check (read speed
1004 listed first, then write speed), and the estimated completion percentage
1005 and time for the running group. It's impossible to estimate runtime of
1006 the following groups (if any).
1008 When fio is done (or interrupted by ctrl-c), it will show the data for
1009 each thread, group of threads, and disks in that order. For each data
1010 direction, the output looks like:
1012 Client1 (g=0): err= 0:
1013 write: io= 32MB, bw= 666KB/s, runt= 50320msec
1014 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
1015 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
1016 bw (KB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
1017 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
1018 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
1019 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1020 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1021 issued r/w: total=0/32768, short=0/0
1022 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
1023 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
1025 The client number is printed, along with the group id and error of that
1026 thread. Below is the io statistics, here for writes. In the order listed,
1029 io= Number of megabytes io performed
1030 bw= Average bandwidth rate
1031 runt= The runtime of that thread
1032 slat= Submission latency (avg being the average, stdev being the
1033 standard deviation). This is the time it took to submit
1034 the io. For sync io, the slat is really the completion
1035 latency, since queue/complete is one operation there. This
1036 value can be in milliseconds or microseconds, fio will choose
1037 the most appropriate base and print that. In the example
1038 above, milliseconds is the best scale.
1039 clat= Completion latency. Same names as slat, this denotes the
1040 time from submission to completion of the io pieces. For
1041 sync io, clat will usually be equal (or very close) to 0,
1042 as the time from submit to complete is basically just
1043 CPU time (io has already been done, see slat explanation).
1044 bw= Bandwidth. Same names as the xlat stats, but also includes
1045 an approximate percentage of total aggregate bandwidth
1046 this thread received in this group. This last value is
1047 only really useful if the threads in this group are on the
1048 same disk, since they are then competing for disk access.
1049 cpu= CPU usage. User and system time, along with the number
1050 of context switches this thread went through, usage of
1051 system and user time, and finally the number of major
1052 and minor page faults.
1053 IO depths= The distribution of io depths over the job life time. The
1054 numbers are divided into powers of 2, so for example the
1055 16= entries includes depths up to that value but higher
1056 than the previous entry. In other words, it covers the
1057 range from 16 to 31.
1058 IO submit= How many pieces of IO were submitting in a single submit
1059 call. Each entry denotes that amount and below, until
1060 the previous entry - eg, 8=100% mean that we submitted
1061 anywhere in between 5-8 ios per submit call.
1062 IO complete= Like the above submit number, but for completions instead.
1063 IO issued= The number of read/write requests issued, and how many
1065 IO latencies= The distribution of IO completion latencies. This is the
1066 time from when IO leaves fio and when it gets completed.
1067 The numbers follow the same pattern as the IO depths,
1068 meaning that 2=1.6% means that 1.6% of the IO completed
1069 within 2 msecs, 20=12.8% means that 12.8% of the IO
1070 took more than 10 msecs, but less than (or equal to) 20 msecs.
1072 After each client has been listed, the group statistics are printed. They
1073 will look like this:
1075 Run status group 0 (all jobs):
1076 READ: io=64MB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
1077 WRITE: io=64MB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
1079 For each data direction, it prints:
1081 io= Number of megabytes io performed.
1082 aggrb= Aggregate bandwidth of threads in this group.
1083 minb= The minimum average bandwidth a thread saw.
1084 maxb= The maximum average bandwidth a thread saw.
1085 mint= The smallest runtime of the threads in that group.
1086 maxt= The longest runtime of the threads in that group.
1088 And finally, the disk statistics are printed. They will look like this:
1090 Disk stats (read/write):
1091 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
1093 Each value is printed for both reads and writes, with reads first. The
1096 ios= Number of ios performed by all groups.
1097 merge= Number of merges io the io scheduler.
1098 ticks= Number of ticks we kept the disk busy.
1099 io_queue= Total time spent in the disk queue.
1100 util= The disk utilization. A value of 100% means we kept the disk
1101 busy constantly, 50% would be a disk idling half of the time.
1107 For scripted usage where you typically want to generate tables or graphs
1108 of the results, fio can output the results in a semicolon separated format.
1109 The format is one long line of values, such as:
1111 client1;0;0;1906777;1090804;1790;0;0;0.000000;0.000000;0;0;0.000000;0.000000;929380;1152890;25.510151%;1078276.333333;128948.113404;0;0;0;0;0;0.000000;0.000000;0;0;0.000000;0.000000;0;0;0.000000%;0.000000;0.000000;100.000000%;0.000000%;324;100.0%;0.0%;0.0%;0.0%;0.0%;0.0%;0.0%;100.0%;0.0%;0.0%;0.0%;0.0%;0.0%
1112 ;0.0%;0.0%;0.0%;0.0%;0.0%
1114 To enable terse output, use the --minimal command line option.
1116 Split up, the format is as follows:
1118 jobname, groupid, error
1120 KB IO, bandwidth (KB/sec), runtime (msec)
1121 Submission latency: min, max, mean, deviation
1122 Completion latency: min, max, mean, deviation
1123 Bw: min, max, aggregate percentage of total, mean, deviation
1125 KB IO, bandwidth (KB/sec), runtime (msec)
1126 Submission latency: min, max, mean, deviation
1127 Completion latency: min, max, mean, deviation
1128 Bw: min, max, aggregate percentage of total, mean, deviation
1129 CPU usage: user, system, context switches, major faults, minor faults
1130 IO depths: <=1, 2, 4, 8, 16, 32, >=64
1131 IO latencies: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, >=2000
projects / fio.git / blob