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 lets look at a really simple job file that define to threads, each
115 randomly reading from a 128MiB 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 Lets look at an example that have 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 32KiB and define numjobs to 4 to
154 fork 4 identical jobs. The result is 4 processes each randomly writing
155 to their own 64MiB 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 ships with a few example job files, you can also look there for
165 5.0 Detailed list of parameters
166 -------------------------------
168 This section describes in details each parameter associated with a job.
169 Some parameters take an option of a given type, such as an integer or
170 a string. The following types are used:
172 str String. This is a sequence of alpha characters.
173 int Integer. A whole number value, can be negative. If prefixed with
174 0x, the integer is assumed to be of base 16 (hexidecimal).
175 siint SI integer. A whole number value, which may contain a postfix
176 describing the base of the number. Accepted postfixes are k/m/g,
177 meaning kilo, mega, and giga. So if you want to specify 4096,
178 you could either write out '4096' or just give 4k. The postfixes
179 signify base 2 values, so 1024 is 1k and 1024k is 1m and so on.
180 If the option accepts an upper and lower range, use a colon ':'
181 or minus '-' to seperate such values. See irange.
182 bool Boolean. Usually parsed as an integer, however only defined for
183 true and false (1 and 0).
184 irange Integer range with postfix. Allows value range to be given, such
185 as 1024-4096. A colon may also be used as the seperator, eg
186 1k:4k. If the option allows two sets of ranges, they can be
187 specified with a ',' or '/' delimiter: 1k-4k/8k-32k. Also see
190 With the above in mind, here follows the complete list of fio job
193 name=str ASCII name of the job. This may be used to override the
194 name printed by fio for this job. Otherwise the job
195 name is used. On the command line this parameter has the
196 special purpose of also signaling the start of a new
199 description=str Text description of the job. Doesn't do anything except
200 dump this text description when this job is run. It's
203 directory=str Prefix filenames with this directory. Used to places files
204 in a different location than "./".
206 filename=str Fio normally makes up a filename based on the job name,
207 thread number, and file number. If you want to share
208 files between threads in a job or several jobs, specify
209 a filename for each of them to override the default. If
210 the ioengine used is 'net', the filename is the host and
211 port to connect to in the format of =host/port. If the
212 ioengine is file based, you can specify a number of files
213 by seperating the names with a ':' colon. So if you wanted
214 a job to open /dev/sda and /dev/sdb as the two working files,
215 you would use filename=/dev/sda:/dev/sdb. '-' is a reserved
216 name, meaning stdin or stdout. Which of the two depends
217 on the read/write direction set.
219 opendir=str Tell fio to recursively add any file it can find in this
220 directory and down the file system tree.
222 lockfile=bool If set, fio will lock a file internally before doing IO to it.
223 This makes it safe to share file descriptors across fio
224 jobs that run at the same time.
226 lockfile_batch=int Acquiring a semaphore can be quite expensive, so
227 allow a process to complete this number of IOs before releasing
228 the semaphore again. Defaults to 1.
231 rw=str Type of io pattern. Accepted values are:
233 read Sequential reads
234 write Sequential writes
235 randwrite Random writes
236 randread Random reads
237 rw Sequential mixed reads and writes
238 randrw Random mixed reads and writes
240 For the mixed io types, the default is to split them 50/50.
241 For certain types of io the result may still be skewed a bit,
242 since the speed may be different. It is possible to specify
243 a number of IO's to do before getting a new offset - this
244 is only useful for random IO, where fio would normally
245 generate a new random offset for every IO. If you append
246 eg 8 to randread, you would get a new random offset for
247 every 8 IO's. The result would be a seek for only every 8
248 IO's, instead of for every IO. Use rw=randread:8 to specify
251 randrepeat=bool For random IO workloads, seed the generator in a predictable
252 way so that results are repeatable across repetitions.
254 fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
255 on what IO patterns it is likely to issue. Sometimes you
256 want to test specific IO patterns without telling the
257 kernel about it, in which case you can disable this option.
258 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
259 IO and POSIX_FADV_RANDOM for random IO.
261 size=siint The total size of file io for this job. Fio will run until
262 this many bytes has been transferred, unless runtime is
263 limited by other options (such as 'runtime', for instance).
264 Unless specific nr_files and filesize options are given,
265 fio will divide this size between the available files
266 specified by the job.
268 filesize=siint Individual file sizes. May be a range, in which case fio
269 will select sizes for files at random within the given range
270 and limited to 'size' in total (if that is given). If not
271 given, each created file is the same size.
273 fill_device=bool Sets size to something really large and waits for ENOSPC (no
274 space left on device) as the terminating condition. Only makes
275 sense with sequential write.
278 bs=siint The block size used for the io units. Defaults to 4k. Values
279 can be given for both read and writes. If a single siint is
280 given, it will apply to both. If a second siint is specified
281 after a comma, it will apply to writes only. In other words,
282 the format is either bs=read_and_write or bs=read,write.
283 bs=4k,8k will thus use 4k blocks for reads, and 8k blocks
284 for writes. If you only wish to set the write size, you
285 can do so by passing an empty read size - bs=,8k will set
286 8k for writes and leave the read default value.
288 blocksize_range=irange
289 bsrange=irange Instead of giving a single block size, specify a range
290 and fio will mix the issued io block sizes. The issued
291 io unit will always be a multiple of the minimum value
292 given (also see bs_unaligned). Applies to both reads and
293 writes, however a second range can be given after a comma.
296 bssplit=str Sometimes you want even finer grained control of the
297 block sizes issued, not just an even split between them.
298 This option allows you to weight various block sizes,
299 so that you are able to define a specific amount of
300 block sizes issued. The format for this option is:
302 bssplit=blocksize/percentage:blocksize/percentage
304 for as many block sizes as needed. So if you want to define
305 a workload that has 50% 64k blocks, 10% 4k blocks, and
306 40% 32k blocks, you would write:
308 bssplit=4k/10:64k/50:32k/40
310 Ordering does not matter. If the percentage is left blank,
311 fio will fill in the remaining values evenly. So a bssplit
312 option like this one:
314 bssplit=4k/50:1k/:32k/
316 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
317 always add up to 100, if bssplit is given a range that adds
318 up to more, it will error out.
321 bs_unaligned If this option is given, any byte size value within bsrange
322 may be used as a block range. This typically wont work with
323 direct IO, as that normally requires sector alignment.
325 zero_buffers If this option is given, fio will init the IO buffers to
326 all zeroes. The default is to fill them with random data.
328 nrfiles=int Number of files to use for this job. Defaults to 1.
330 openfiles=int Number of files to keep open at the same time. Defaults to
331 the same as nrfiles, can be set smaller to limit the number
334 file_service_type=str Defines how fio decides which file from a job to
335 service next. The following types are defined:
337 random Just choose a file at random.
339 roundrobin Round robin over open files. This
342 The string can have a number appended, indicating how
343 often to switch to a new file. So if option random:4 is
344 given, fio will switch to a new random file after 4 ios
347 ioengine=str Defines how the job issues io to the file. The following
350 sync Basic read(2) or write(2) io. lseek(2) is
351 used to position the io location.
353 psync Basic pread(2) or pwrite(2) io.
355 vsync Basic readv(2) or writev(2) IO.
357 libaio Linux native asynchronous io.
359 posixaio glibc posix asynchronous io.
361 mmap File is memory mapped and data copied
362 to/from using memcpy(3).
364 splice splice(2) is used to transfer the data and
365 vmsplice(2) to transfer data from user
368 syslet-rw Use the syslet system calls to make
369 regular read/write async.
371 sg SCSI generic sg v3 io. May either be
372 synchronous using the SG_IO ioctl, or if
373 the target is an sg character device
374 we use read(2) and write(2) for asynchronous
377 null Doesn't transfer any data, just pretends
378 to. This is mainly used to exercise fio
379 itself and for debugging/testing purposes.
381 net Transfer over the network to given host:port.
382 'filename' must be set appropriately to
383 filename=host/port regardless of send
384 or receive, if the latter only the port
387 netsplice Like net, but uses splice/vmsplice to
388 map data and send/receive.
390 cpuio Doesn't transfer any data, but burns CPU
391 cycles according to the cpuload= and
392 cpucycle= options. Setting cpuload=85
393 will cause that job to do nothing but burn
394 85% of the CPU. In case of SMP machines,
395 use numjobs=<no_of_cpu> to get desired CPU
396 usage, as the cpuload only loads a single
397 CPU at the desired rate.
399 guasi The GUASI IO engine is the Generic Userspace
400 Asyncronous Syscall Interface approach
403 http://www.xmailserver.org/guasi-lib.html
405 for more info on GUASI.
407 external Prefix to specify loading an external
408 IO engine object file. Append the engine
409 filename, eg ioengine=external:/tmp/foo.o
410 to load ioengine foo.o in /tmp.
412 iodepth=int This defines how many io units to keep in flight against
413 the file. The default is 1 for each file defined in this
414 job, can be overridden with a larger value for higher
417 iodepth_batch=int This defines how many pieces of IO to submit at once.
418 It defaults to 1 which means that we submit each IO
419 as soon as it is available, but can be raised to submit
420 bigger batches of IO at the time.
422 iodepth_low=int The low water mark indicating when to start filling
423 the queue again. Defaults to the same as iodepth, meaning
424 that fio will attempt to keep the queue full at all times.
425 If iodepth is set to eg 16 and iodepth_low is set to 4, then
426 after fio has filled the queue of 16 requests, it will let
427 the depth drain down to 4 before starting to fill it again.
429 direct=bool If value is true, use non-buffered io. This is usually
432 buffered=bool If value is true, use buffered io. This is the opposite
433 of the 'direct' option. Defaults to true.
435 offset=siint Start io at the given offset in the file. The data before
436 the given offset will not be touched. This effectively
437 caps the file size at real_size - offset.
439 fsync=int If writing to a file, issue a sync of the dirty data
440 for every number of blocks given. For example, if you give
441 32 as a parameter, fio will sync the file for every 32
442 writes issued. If fio is using non-buffered io, we may
443 not sync the file. The exception is the sg io engine, which
444 synchronizes the disk cache anyway.
446 overwrite=bool If writing to a file, setup the file first and do overwrites.
448 end_fsync=bool If true, fsync file contents when the job exits.
450 fsync_on_close=bool If true, fio will fsync() a dirty file on close.
451 This differs from end_fsync in that it will happen on every
452 file close, not just at the end of the job.
454 rwmixcycle=int Value in milliseconds describing how often to switch between
455 reads and writes for a mixed workload. The default is
458 rwmixread=int How large a percentage of the mix should be reads.
460 rwmixwrite=int How large a percentage of the mix should be writes. If both
461 rwmixread and rwmixwrite is given and the values do not add
462 up to 100%, the latter of the two will be used to override
465 norandommap Normally fio will cover every block of the file when doing
466 random IO. If this option is given, fio will just get a
467 new random offset without looking at past io history. This
468 means that some blocks may not be read or written, and that
469 some blocks may be read/written more than once. This option
470 is mutually exclusive with verify= for that reason, since
471 fio doesn't track potential block rewrites which may alter
472 the calculated checksum for that block.
474 nice=int Run the job with the given nice value. See man nice(2).
476 prio=int Set the io priority value of this job. Linux limits us to
477 a positive value between 0 and 7, with 0 being the highest.
480 prioclass=int Set the io priority class. See man ionice(1).
482 thinktime=int Stall the job x microseconds after an io has completed before
483 issuing the next. May be used to simulate processing being
484 done by an application. See thinktime_blocks and
488 Only valid if thinktime is set - pretend to spend CPU time
489 doing something with the data received, before falling back
490 to sleeping for the rest of the period specified by
494 Only valid if thinktime is set - control how many blocks
495 to issue, before waiting 'thinktime' usecs. If not set,
496 defaults to 1 which will make fio wait 'thinktime' usecs
499 rate=int Cap the bandwidth used by this job to this number of KiB/sec.
501 ratemin=int Tell fio to do whatever it can to maintain at least this
502 bandwidth. Failing to meet this requirement, will cause
505 rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
506 as rate, just specified independently of bandwidth. If the
507 job is given a block size range instead of a fixed value,
508 the smallest block size is used as the metric.
510 rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
513 ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
516 cpumask=int Set the CPU affinity of this job. The parameter given is a
517 bitmask of allowed CPU's the job may run on. So if you want
518 the allowed CPUs to be 1 and 5, you would pass the decimal
519 value of (1 << 1 | 1 << 5), or 34. See man
520 sched_setaffinity(2). This may not work on all supported
521 operating systems or kernel versions.
523 cpus_allowed=str Controls the same options as cpumask, but it allows a text
524 setting of the permitted CPUs instead. So to use CPUs 1 and
525 5, you would specify cpus_allowed=1,5.
527 startdelay=int Start this job the specified number of seconds after fio
528 has started. Only useful if the job file contains several
529 jobs, and you want to delay starting some jobs to a certain
532 runtime=int Tell fio to terminate processing after the specified number
533 of seconds. It can be quite hard to determine for how long
534 a specified job will run, so this parameter is handy to
535 cap the total runtime to a given time.
537 time_based If set, fio will run for the duration of the runtime
538 specified even if the file(s) are completey read or
539 written. It will simply loop over the same workload
540 as many times as the runtime allows.
542 invalidate=bool Invalidate the buffer/page cache parts for this file prior
543 to starting io. Defaults to true.
545 sync=bool Use sync io for buffered writes. For the majority of the
546 io engines, this means using O_SYNC.
549 mem=str Fio can use various types of memory as the io unit buffer.
550 The allowed values are:
552 malloc Use memory from malloc(3) as the buffers.
554 shm Use shared memory as the buffers. Allocated
557 shmhuge Same as shm, but use huge pages as backing.
559 mmap Use mmap to allocate buffers. May either be
560 anonymous memory, or can be file backed if
561 a filename is given after the option. The
562 format is mem=mmap:/path/to/file.
564 mmaphuge Use a memory mapped huge file as the buffer
565 backing. Append filename after mmaphuge, ala
566 mem=mmaphuge:/hugetlbfs/file
568 The area allocated is a function of the maximum allowed
569 bs size for the job, multiplied by the io depth given. Note
570 that for shmhuge and mmaphuge to work, the system must have
571 free huge pages allocated. This can normally be checked
572 and set by reading/writing /proc/sys/vm/nr_hugepages on a
573 Linux system. Fio assumes a huge page is 4MiB in size. So
574 to calculate the number of huge pages you need for a given
575 job file, add up the io depth of all jobs (normally one unless
576 iodepth= is used) and multiply by the maximum bs set. Then
577 divide that number by the huge page size. You can see the
578 size of the huge pages in /proc/meminfo. If no huge pages
579 are allocated by having a non-zero number in nr_hugepages,
580 using mmaphuge or shmhuge will fail. Also see hugepage-size.
582 mmaphuge also needs to have hugetlbfs mounted and the file
583 location should point there. So if it's mounted in /huge,
584 you would use mem=mmaphuge:/huge/somefile.
587 Defines the size of a huge page. Must at least be equal
588 to the system setting, see /proc/meminfo. Defaults to 4MiB.
589 Should probably always be a multiple of megabytes, so using
590 hugepage-size=Xm is the preferred way to set this to avoid
591 setting a non-pow-2 bad value.
593 exitall When one job finishes, terminate the rest. The default is
594 to wait for each job to finish, sometimes that is not the
597 bwavgtime=int Average the calculated bandwidth over the given time. Value
598 is specified in milliseconds.
600 create_serialize=bool If true, serialize the file creating for the jobs.
601 This may be handy to avoid interleaving of data
602 files, which may greatly depend on the filesystem
603 used and even the number of processors in the system.
605 create_fsync=bool fsync the data file after creation. This is the
608 unlink=bool Unlink the job files when done. Not the default, as repeated
609 runs of that job would then waste time recreating the fileset
612 loops=int Run the specified number of iterations of this job. Used
613 to repeat the same workload a given number of times. Defaults
616 do_verify=bool Run the verify phase after a write phase. Only makes sense if
617 verify is set. Defaults to 1.
619 verify=str If writing to a file, fio can verify the file contents
620 after each iteration of the job. The allowed values are:
622 md5 Use an md5 sum of the data area and store
623 it in the header of each block.
625 crc64 Use an experimental crc64 sum of the data
626 area and store it in the header of each
629 crc32 Use a crc32 sum of the data area and store
630 it in the header of each block.
632 crc16 Use a crc16 sum of the data area and store
633 it in the header of each block.
635 crc7 Use a crc7 sum of the data area and store
636 it in the header of each block.
638 sha512 Use sha512 as the checksum function.
640 sha256 Use sha256 as the checksum function.
642 meta Write extra information about each io
643 (timestamp, block number etc.). The block
646 null Only pretend to verify. Useful for testing
647 internals with ioengine=null, not for much
650 This option can be used for repeated burn-in tests of a
651 system to make sure that the written data is also
654 verifysort=bool If set, fio will sort written verify blocks when it deems
655 it faster to read them back in a sorted manner. This is
656 often the case when overwriting an existing file, since
657 the blocks are already laid out in the file system. You
658 can ignore this option unless doing huge amounts of really
659 fast IO where the red-black tree sorting CPU time becomes
662 verify_offset=siint Swap the verification header with data somewhere else
663 in the block before writing. Its swapped back before
666 verify_interval=siint Write the verification header at a finer granularity
667 than the blocksize. It will be written for chunks the
668 size of header_interval. blocksize should divide this
671 verify_pattern=int If set, fio will fill the io buffers with this
672 pattern. Fio defaults to filling with totally random
673 bytes, but sometimes it's interesting to fill with a known
674 pattern for io verification purposes. Depending on the
675 width of the pattern, fio will fill 1/2/3/4 bytes of the
676 buffer at the time. The verify_pattern cannot be larger than
679 verify_fatal=bool Normally fio will keep checking the entire contents
680 before quitting on a block verification failure. If this
681 option is set, fio will exit the job on the first observed
684 stonewall Wait for preceeding jobs in the job file to exit, before
685 starting this one. Can be used to insert serialization
686 points in the job file. A stone wall also implies starting
687 a new reporting group.
689 new_group Start a new reporting group. If this option isn't given,
690 jobs in a file will be part of the same reporting group
691 unless seperated by a stone wall (or if it's a group
692 by itself, with the numjobs option).
694 numjobs=int Create the specified number of clones of this job. May be
695 used to setup a larger number of threads/processes doing
696 the same thing. We regard that grouping of jobs as a
699 group_reporting If 'numjobs' is set, it may be interesting to display
700 statistics for the group as a whole instead of for each
701 individual job. This is especially true of 'numjobs' is
702 large, looking at individual thread/process output quickly
703 becomes unwieldy. If 'group_reporting' is specified, fio
704 will show the final report per-group instead of per-job.
706 thread fio defaults to forking jobs, however if this option is
707 given, fio will use pthread_create(3) to create threads
710 zonesize=siint Divide a file into zones of the specified size. See zoneskip.
712 zoneskip=siint Skip the specified number of bytes when zonesize data has
713 been read. The two zone options can be used to only do
714 io on zones of a file.
716 write_iolog=str Write the issued io patterns to the specified file. See
719 read_iolog=str Open an iolog with the specified file name and replay the
720 io patterns it contains. This can be used to store a
721 workload and replay it sometime later. The iolog given
722 may also be a blktrace binary file, which allows fio
723 to replay a workload captured by blktrace. See blktrace
724 for how to capture such logging data. For blktrace replay,
725 the file needs to be turned into a blkparse binary data
726 file first (blktrace <device> -d file_for_fio.bin).
728 write_bw_log If given, write a bandwidth log of the jobs in this job
729 file. Can be used to store data of the bandwidth of the
730 jobs in their lifetime. The included fio_generate_plots
731 script uses gnuplot to turn these text files into nice
734 write_lat_log Same as write_bw_log, except that this option stores io
735 completion latencies instead.
737 lockmem=siint Pin down the specified amount of memory with mlock(2). Can
738 potentially be used instead of removing memory or booting
739 with less memory to simulate a smaller amount of memory.
741 exec_prerun=str Before running this job, issue the command specified
744 exec_postrun=str After the job completes, issue the command specified
747 ioscheduler=str Attempt to switch the device hosting the file to the specified
748 io scheduler before running.
750 cpuload=int If the job is a CPU cycle eater, attempt to use the specified
751 percentage of CPU cycles.
753 cpuchunks=int If the job is a CPU cycle eater, split the load into
754 cycles of the given time. In milliseconds.
756 disk_util=bool Generate disk utilization statistics, if the platform
757 supports it. Defaults to on.
760 6.0 Interpreting the output
761 ---------------------------
763 fio spits out a lot of output. While running, fio will display the
764 status of the jobs created. An example of that would be:
766 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
768 The characters inside the square brackets denote the current status of
769 each thread. The possible values (in typical life cycle order) are:
773 P Thread setup, but not started.
775 I Thread initialized, waiting.
776 R Running, doing sequential reads.
777 r Running, doing random reads.
778 W Running, doing sequential writes.
779 w Running, doing random writes.
780 M Running, doing mixed sequential reads/writes.
781 m Running, doing mixed random reads/writes.
782 F Running, currently waiting for fsync()
783 V Running, doing verification of written data.
784 E Thread exited, not reaped by main thread yet.
787 The other values are fairly self explanatory - number of threads
788 currently running and doing io, rate of io since last check (read speed
789 listed first, then write speed), and the estimated completion percentage
790 and time for the running group. It's impossible to estimate runtime of
791 the following groups (if any).
793 When fio is done (or interrupted by ctrl-c), it will show the data for
794 each thread, group of threads, and disks in that order. For each data
795 direction, the output looks like:
797 Client1 (g=0): err= 0:
798 write: io= 32MiB, bw= 666KiB/s, runt= 50320msec
799 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
800 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
801 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
802 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
803 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
804 issued r/w: total=0/32768, short=0/0
805 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
806 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
808 The client number is printed, along with the group id and error of that
809 thread. Below is the io statistics, here for writes. In the order listed,
812 io= Number of megabytes io performed
813 bw= Average bandwidth rate
814 runt= The runtime of that thread
815 slat= Submission latency (avg being the average, stdev being the
816 standard deviation). This is the time it took to submit
817 the io. For sync io, the slat is really the completion
818 latency, since queue/complete is one operation there. This
819 value can be in miliseconds or microseconds, fio will choose
820 the most appropriate base and print that. In the example
821 above, miliseconds is the best scale.
822 clat= Completion latency. Same names as slat, this denotes the
823 time from submission to completion of the io pieces. For
824 sync io, clat will usually be equal (or very close) to 0,
825 as the time from submit to complete is basically just
826 CPU time (io has already been done, see slat explanation).
827 bw= Bandwidth. Same names as the xlat stats, but also includes
828 an approximate percentage of total aggregate bandwidth
829 this thread received in this group. This last value is
830 only really useful if the threads in this group are on the
831 same disk, since they are then competing for disk access.
832 cpu= CPU usage. User and system time, along with the number
833 of context switches this thread went through, usage of
834 system and user time, and finally the number of major
835 and minor page faults.
836 IO depths= The distribution of io depths over the job life time. The
837 numbers are divided into powers of 2, so for example the
838 16= entries includes depths up to that value but higher
839 than the previous entry. In other words, it covers the
841 IO issued= The number of read/write requests issued, and how many
843 IO latencies= The distribution of IO completion latencies. This is the
844 time from when IO leaves fio and when it gets completed.
845 The numbers follow the same pattern as the IO depths,
846 meaning that 2=1.6% means that 1.6% of the IO completed
847 within 2 msecs, 20=12.8% means that 12.8% of the IO
848 took more than 10 msecs, but less than (or equal to) 20 msecs.
850 After each client has been listed, the group statistics are printed. They
853 Run status group 0 (all jobs):
854 READ: io=64MiB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
855 WRITE: io=64MiB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
857 For each data direction, it prints:
859 io= Number of megabytes io performed.
860 aggrb= Aggregate bandwidth of threads in this group.
861 minb= The minimum average bandwidth a thread saw.
862 maxb= The maximum average bandwidth a thread saw.
863 mint= The smallest runtime of the threads in that group.
864 maxt= The longest runtime of the threads in that group.
866 And finally, the disk statistics are printed. They will look like this:
868 Disk stats (read/write):
869 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
871 Each value is printed for both reads and writes, with reads first. The
874 ios= Number of ios performed by all groups.
875 merge= Number of merges io the io scheduler.
876 ticks= Number of ticks we kept the disk busy.
877 io_queue= Total time spent in the disk queue.
878 util= The disk utilization. A value of 100% means we kept the disk
879 busy constantly, 50% would be a disk idling half of the time.
885 For scripted usage where you typically want to generate tables or graphs
886 of the results, fio can output the results in a semicolon separated format.
887 The format is one long line of values, such as:
889 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%
890 ;0.0%;0.0%;0.0%;0.0%;0.0%
892 Split up, the format is as follows:
894 jobname, groupid, error
896 KiB IO, bandwidth (KiB/sec), runtime (msec)
897 Submission latency: min, max, mean, deviation
898 Completion latency: min, max, mean, deviation
899 Bw: min, max, aggregate percentage of total, mean, deviation
901 KiB IO, bandwidth (KiB/sec), runtime (msec)
902 Submission latency: min, max, mean, deviation
903 Completion latency: min, max, mean, deviation
904 Bw: min, max, aggregate percentage of total, mean, deviation
905 CPU usage: user, system, context switches, major faults, minor faults
906 IO depths: <=1, 2, 4, 8, 16, 32, >=64
907 IO latencies: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, >=2000