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 (hexadecimal).
175 time Integer with possible time postfix. In seconds unless otherwise
176 specified, use eg 10m for 10 minutes. Accepts s/m/h for seconds,
178 siint SI integer. A whole number value, which may contain a postfix
179 describing the base of the number. Accepted postfixes are k/m/g,
180 meaning kilo, mega, and giga. So if you want to specify 4096,
181 you could either write out '4096' or just give 4k. The postfixes
182 signify base 2 values, so 1024 is 1k and 1024k is 1m and so on.
183 If the option accepts an upper and lower range, use a colon ':'
184 or minus '-' to separate such values. See irange.
185 bool Boolean. Usually parsed as an integer, however only defined for
186 true and false (1 and 0).
187 irange Integer range with postfix. Allows value range to be given, such
188 as 1024-4096. A colon may also be used as the separator, eg
189 1k:4k. If the option allows two sets of ranges, they can be
190 specified with a ',' or '/' delimiter: 1k-4k/8k-32k. Also see
193 With the above in mind, here follows the complete list of fio job
196 name=str ASCII name of the job. This may be used to override the
197 name printed by fio for this job. Otherwise the job
198 name is used. On the command line this parameter has the
199 special purpose of also signaling the start of a new
202 description=str Text description of the job. Doesn't do anything except
203 dump this text description when this job is run. It's
206 directory=str Prefix filenames with this directory. Used to places files
207 in a different location than "./".
209 filename=str Fio normally makes up a filename based on the job name,
210 thread number, and file number. If you want to share
211 files between threads in a job or several jobs, specify
212 a filename for each of them to override the default. If
213 the ioengine used is 'net', the filename is the host and
214 port to connect to in the format of =host/port. If the
215 ioengine is file based, you can specify a number of files
216 by separating the names with a ':' colon. So if you wanted
217 a job to open /dev/sda and /dev/sdb as the two working files,
218 you would use filename=/dev/sda:/dev/sdb. '-' is a reserved
219 name, meaning stdin or stdout. Which of the two depends
220 on the read/write direction set.
222 opendir=str Tell fio to recursively add any file it can find in this
223 directory and down the file system tree.
225 lockfile=str Fio defaults to not doing any locking files before it does
226 IO to them. If a file or file descriptor is shared, fio
227 can serialize IO to that file to make the end result
228 consistent. This is usual for emulating real workloads that
229 share files. The lock modes are:
231 none No locking. The default.
232 exclusive Only one thread/process may do IO,
233 excluding all others.
234 readwrite Read-write locking on the file. Many
235 readers may access the file at the
236 same time, but writes get exclusive
239 The option may be post-fixed with a lock batch number. If
240 set, then each thread/process may do that amount of IOs to
241 the file before giving up the lock. Since lock acquisition is
242 expensive, batching the lock/unlocks will speed up IO.
245 rw=str Type of io pattern. Accepted values are:
247 read Sequential reads
248 write Sequential writes
249 randwrite Random writes
250 randread Random reads
251 rw Sequential mixed reads and writes
252 randrw Random mixed reads and writes
254 For the mixed io types, the default is to split them 50/50.
255 For certain types of io the result may still be skewed a bit,
256 since the speed may be different. It is possible to specify
257 a number of IO's to do before getting a new offset - this
258 is only useful for random IO, where fio would normally
259 generate a new random offset for every IO. If you append
260 eg 8 to randread, you would get a new random offset for
261 every 8 IO's. The result would be a seek for only every 8
262 IO's, instead of for every IO. Use rw=randread:8 to specify
265 randrepeat=bool For random IO workloads, seed the generator in a predictable
266 way so that results are repeatable across repetitions.
268 fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
269 on what IO patterns it is likely to issue. Sometimes you
270 want to test specific IO patterns without telling the
271 kernel about it, in which case you can disable this option.
272 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
273 IO and POSIX_FADV_RANDOM for random IO.
275 size=siint The total size of file io for this job. Fio will run until
276 this many bytes has been transferred, unless runtime is
277 limited by other options (such as 'runtime', for instance).
278 Unless specific nr_files and filesize options are given,
279 fio will divide this size between the available files
280 specified by the job.
282 filesize=siint Individual file sizes. May be a range, in which case fio
283 will select sizes for files at random within the given range
284 and limited to 'size' in total (if that is given). If not
285 given, each created file is the same size.
287 fill_device=bool Sets size to something really large and waits for ENOSPC (no
288 space left on device) as the terminating condition. Only makes
289 sense with sequential write.
292 bs=siint The block size used for the io units. Defaults to 4k. Values
293 can be given for both read and writes. If a single siint is
294 given, it will apply to both. If a second siint is specified
295 after a comma, it will apply to writes only. In other words,
296 the format is either bs=read_and_write or bs=read,write.
297 bs=4k,8k will thus use 4k blocks for reads, and 8k blocks
298 for writes. If you only wish to set the write size, you
299 can do so by passing an empty read size - bs=,8k will set
300 8k for writes and leave the read default value.
302 blocksize_range=irange
303 bsrange=irange Instead of giving a single block size, specify a range
304 and fio will mix the issued io block sizes. The issued
305 io unit will always be a multiple of the minimum value
306 given (also see bs_unaligned). Applies to both reads and
307 writes, however a second range can be given after a comma.
310 bssplit=str Sometimes you want even finer grained control of the
311 block sizes issued, not just an even split between them.
312 This option allows you to weight various block sizes,
313 so that you are able to define a specific amount of
314 block sizes issued. The format for this option is:
316 bssplit=blocksize/percentage:blocksize/percentage
318 for as many block sizes as needed. So if you want to define
319 a workload that has 50% 64k blocks, 10% 4k blocks, and
320 40% 32k blocks, you would write:
322 bssplit=4k/10:64k/50:32k/40
324 Ordering does not matter. If the percentage is left blank,
325 fio will fill in the remaining values evenly. So a bssplit
326 option like this one:
328 bssplit=4k/50:1k/:32k/
330 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
331 always add up to 100, if bssplit is given a range that adds
332 up to more, it will error out.
335 bs_unaligned If this option is given, any byte size value within bsrange
336 may be used as a block range. This typically wont work with
337 direct IO, as that normally requires sector alignment.
339 zero_buffers If this option is given, fio will init the IO buffers to
340 all zeroes. The default is to fill them with random data.
342 refill_buffers If this option is given, fio will refill the IO buffers
343 on every submit. The default is to only fill it at init
344 time and reuse that data. Only makes sense if zero_buffers
345 isn't specified, naturally. If data verification is enabled,
346 refill_buffers is also automatically enabled.
348 nrfiles=int Number of files to use for this job. Defaults to 1.
350 openfiles=int Number of files to keep open at the same time. Defaults to
351 the same as nrfiles, can be set smaller to limit the number
354 file_service_type=str Defines how fio decides which file from a job to
355 service next. The following types are defined:
357 random Just choose a file at random.
359 roundrobin Round robin over open files. This
362 The string can have a number appended, indicating how
363 often to switch to a new file. So if option random:4 is
364 given, fio will switch to a new random file after 4 ios
367 ioengine=str Defines how the job issues io to the file. The following
370 sync Basic read(2) or write(2) io. lseek(2) is
371 used to position the io location.
373 psync Basic pread(2) or pwrite(2) io.
375 vsync Basic readv(2) or writev(2) IO.
377 libaio Linux native asynchronous io.
379 posixaio glibc posix asynchronous io.
381 solarisaio Solaris native asynchronous io.
383 mmap File is memory mapped and data copied
384 to/from using memcpy(3).
386 splice splice(2) is used to transfer the data and
387 vmsplice(2) to transfer data from user
390 syslet-rw Use the syslet system calls to make
391 regular read/write async.
393 sg SCSI generic sg v3 io. May either be
394 synchronous using the SG_IO ioctl, or if
395 the target is an sg character device
396 we use read(2) and write(2) for asynchronous
399 null Doesn't transfer any data, just pretends
400 to. This is mainly used to exercise fio
401 itself and for debugging/testing purposes.
403 net Transfer over the network to given host:port.
404 'filename' must be set appropriately to
405 filename=host/port regardless of send
406 or receive, if the latter only the port
409 netsplice Like net, but uses splice/vmsplice to
410 map data and send/receive.
412 cpuio Doesn't transfer any data, but burns CPU
413 cycles according to the cpuload= and
414 cpucycle= options. Setting cpuload=85
415 will cause that job to do nothing but burn
416 85% of the CPU. In case of SMP machines,
417 use numjobs=<no_of_cpu> to get desired CPU
418 usage, as the cpuload only loads a single
419 CPU at the desired rate.
421 guasi The GUASI IO engine is the Generic Userspace
422 Asyncronous Syscall Interface approach
425 http://www.xmailserver.org/guasi-lib.html
427 for more info on GUASI.
429 external Prefix to specify loading an external
430 IO engine object file. Append the engine
431 filename, eg ioengine=external:/tmp/foo.o
432 to load ioengine foo.o in /tmp.
434 iodepth=int This defines how many io units to keep in flight against
435 the file. The default is 1 for each file defined in this
436 job, can be overridden with a larger value for higher
439 iodepth_batch_submit=int
440 iodepth_batch=int This defines how many pieces of IO to submit at once.
441 It defaults to 1 which means that we submit each IO
442 as soon as it is available, but can be raised to submit
443 bigger batches of IO at the time.
445 iodepth_batch_complete=int This defines how many pieces of IO to retrieve
446 at once. It defaults to 1 which means that we'll ask
447 for a minimum of 1 IO in the retrieval process from
448 the kernel. The IO retrieval will go on until we
449 hit the limit set by iodepth_low. If this variable is
450 set to 0, then fio will always check for completed
451 events before queuing more IO. This helps reduce
452 IO latency, at the cost of more retrieval system calls.
454 iodepth_low=int The low water mark indicating when to start filling
455 the queue again. Defaults to the same as iodepth, meaning
456 that fio will attempt to keep the queue full at all times.
457 If iodepth is set to eg 16 and iodepth_low is set to 4, then
458 after fio has filled the queue of 16 requests, it will let
459 the depth drain down to 4 before starting to fill it again.
461 direct=bool If value is true, use non-buffered io. This is usually
464 buffered=bool If value is true, use buffered io. This is the opposite
465 of the 'direct' option. Defaults to true.
467 offset=siint Start io at the given offset in the file. The data before
468 the given offset will not be touched. This effectively
469 caps the file size at real_size - offset.
471 fsync=int If writing to a file, issue a sync of the dirty data
472 for every number of blocks given. For example, if you give
473 32 as a parameter, fio will sync the file for every 32
474 writes issued. If fio is using non-buffered io, we may
475 not sync the file. The exception is the sg io engine, which
476 synchronizes the disk cache anyway.
478 overwrite=bool If true, writes to a file will always overwrite existing
479 data. If the file doesn't already exist, it will be
480 created before the write phase begins. If the file exists
481 and is large enough for the specified write phase, nothing
484 end_fsync=bool If true, fsync file contents when the job exits.
486 fsync_on_close=bool If true, fio will fsync() a dirty file on close.
487 This differs from end_fsync in that it will happen on every
488 file close, not just at the end of the job.
490 rwmixread=int How large a percentage of the mix should be reads.
492 rwmixwrite=int How large a percentage of the mix should be writes. If both
493 rwmixread and rwmixwrite is given and the values do not add
494 up to 100%, the latter of the two will be used to override
497 norandommap Normally fio will cover every block of the file when doing
498 random IO. If this option is given, fio will just get a
499 new random offset without looking at past io history. This
500 means that some blocks may not be read or written, and that
501 some blocks may be read/written more than once. This option
502 is mutually exclusive with verify= for that reason, since
503 fio doesn't track potential block rewrites which may alter
504 the calculated checksum for that block.
506 softrandommap See norandommap. If fio runs with the random block map enabled
507 and it fails to allocate the map, if this option is set it
508 will continue without a random block map. As coverage will
509 not be as complete as with random maps, this option is
512 nice=int Run the job with the given nice value. See man nice(2).
514 prio=int Set the io priority value of this job. Linux limits us to
515 a positive value between 0 and 7, with 0 being the highest.
518 prioclass=int Set the io priority class. See man ionice(1).
520 thinktime=int Stall the job x microseconds after an io has completed before
521 issuing the next. May be used to simulate processing being
522 done by an application. See thinktime_blocks and
526 Only valid if thinktime is set - pretend to spend CPU time
527 doing something with the data received, before falling back
528 to sleeping for the rest of the period specified by
532 Only valid if thinktime is set - control how many blocks
533 to issue, before waiting 'thinktime' usecs. If not set,
534 defaults to 1 which will make fio wait 'thinktime' usecs
537 rate=int Cap the bandwidth used by this job to this number of KiB/sec.
539 ratemin=int Tell fio to do whatever it can to maintain at least this
540 bandwidth. Failing to meet this requirement, will cause
543 rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
544 as rate, just specified independently of bandwidth. If the
545 job is given a block size range instead of a fixed value,
546 the smallest block size is used as the metric.
548 rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
551 ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
554 cpumask=int Set the CPU affinity of this job. The parameter given is a
555 bitmask of allowed CPU's the job may run on. So if you want
556 the allowed CPUs to be 1 and 5, you would pass the decimal
557 value of (1 << 1 | 1 << 5), or 34. See man
558 sched_setaffinity(2). This may not work on all supported
559 operating systems or kernel versions.
561 cpus_allowed=str Controls the same options as cpumask, but it allows a text
562 setting of the permitted CPUs instead. So to use CPUs 1 and
563 5, you would specify cpus_allowed=1,5.
565 startdelay=time Start this job the specified number of seconds after fio
566 has started. Only useful if the job file contains several
567 jobs, and you want to delay starting some jobs to a certain
570 runtime=time Tell fio to terminate processing after the specified number
571 of seconds. It can be quite hard to determine for how long
572 a specified job will run, so this parameter is handy to
573 cap the total runtime to a given time.
575 time_based If set, fio will run for the duration of the runtime
576 specified even if the file(s) are completely read or
577 written. It will simply loop over the same workload
578 as many times as the runtime allows.
580 ramp_time=time If set, fio will run the specified workload for this amount
581 of time before logging any performance numbers. Useful for
582 letting performance settle before logging results, thus
583 minimizing the runtime required for stable results. Note
584 that the ramp_time is considered lead in time for a job,
585 thus it will increase the total runtime if a special timeout
586 or runtime is specified.
588 invalidate=bool Invalidate the buffer/page cache parts for this file prior
589 to starting io. Defaults to true.
591 sync=bool Use sync io for buffered writes. For the majority of the
592 io engines, this means using O_SYNC.
595 mem=str Fio can use various types of memory as the io unit buffer.
596 The allowed values are:
598 malloc Use memory from malloc(3) as the buffers.
600 shm Use shared memory as the buffers. Allocated
603 shmhuge Same as shm, but use huge pages as backing.
605 mmap Use mmap to allocate buffers. May either be
606 anonymous memory, or can be file backed if
607 a filename is given after the option. The
608 format is mem=mmap:/path/to/file.
610 mmaphuge Use a memory mapped huge file as the buffer
611 backing. Append filename after mmaphuge, ala
612 mem=mmaphuge:/hugetlbfs/file
614 The area allocated is a function of the maximum allowed
615 bs size for the job, multiplied by the io depth given. Note
616 that for shmhuge and mmaphuge to work, the system must have
617 free huge pages allocated. This can normally be checked
618 and set by reading/writing /proc/sys/vm/nr_hugepages on a
619 Linux system. Fio assumes a huge page is 4MiB in size. So
620 to calculate the number of huge pages you need for a given
621 job file, add up the io depth of all jobs (normally one unless
622 iodepth= is used) and multiply by the maximum bs set. Then
623 divide that number by the huge page size. You can see the
624 size of the huge pages in /proc/meminfo. If no huge pages
625 are allocated by having a non-zero number in nr_hugepages,
626 using mmaphuge or shmhuge will fail. Also see hugepage-size.
628 mmaphuge also needs to have hugetlbfs mounted and the file
629 location should point there. So if it's mounted in /huge,
630 you would use mem=mmaphuge:/huge/somefile.
633 Defines the size of a huge page. Must at least be equal
634 to the system setting, see /proc/meminfo. Defaults to 4MiB.
635 Should probably always be a multiple of megabytes, so using
636 hugepage-size=Xm is the preferred way to set this to avoid
637 setting a non-pow-2 bad value.
639 exitall When one job finishes, terminate the rest. The default is
640 to wait for each job to finish, sometimes that is not the
643 bwavgtime=int Average the calculated bandwidth over the given time. Value
644 is specified in milliseconds.
646 create_serialize=bool If true, serialize the file creating for the jobs.
647 This may be handy to avoid interleaving of data
648 files, which may greatly depend on the filesystem
649 used and even the number of processors in the system.
651 create_fsync=bool fsync the data file after creation. This is the
654 unlink=bool Unlink the job files when done. Not the default, as repeated
655 runs of that job would then waste time recreating the file
658 loops=int Run the specified number of iterations of this job. Used
659 to repeat the same workload a given number of times. Defaults
662 do_verify=bool Run the verify phase after a write phase. Only makes sense if
663 verify is set. Defaults to 1.
665 verify=str If writing to a file, fio can verify the file contents
666 after each iteration of the job. The allowed values are:
668 md5 Use an md5 sum of the data area and store
669 it in the header of each block.
671 crc64 Use an experimental crc64 sum of the data
672 area and store it in the header of each
675 crc32c Use a crc32c sum of the data area and store
676 it in the header of each block.
678 crc32c-intel Use hardware assisted crc32c calcuation
679 provided on SSE4.2 enabled processors.
681 crc32 Use a crc32 sum of the data area and store
682 it in the header of each block.
684 crc16 Use a crc16 sum of the data area and store
685 it in the header of each block.
687 crc7 Use a crc7 sum of the data area and store
688 it in the header of each block.
690 sha512 Use sha512 as the checksum function.
692 sha256 Use sha256 as the checksum function.
694 meta Write extra information about each io
695 (timestamp, block number etc.). The block
698 null Only pretend to verify. Useful for testing
699 internals with ioengine=null, not for much
702 This option can be used for repeated burn-in tests of a
703 system to make sure that the written data is also
706 verifysort=bool If set, fio will sort written verify blocks when it deems
707 it faster to read them back in a sorted manner. This is
708 often the case when overwriting an existing file, since
709 the blocks are already laid out in the file system. You
710 can ignore this option unless doing huge amounts of really
711 fast IO where the red-black tree sorting CPU time becomes
714 verify_offset=siint Swap the verification header with data somewhere else
715 in the block before writing. Its swapped back before
718 verify_interval=siint Write the verification header at a finer granularity
719 than the blocksize. It will be written for chunks the
720 size of header_interval. blocksize should divide this
723 verify_pattern=int If set, fio will fill the io buffers with this
724 pattern. Fio defaults to filling with totally random
725 bytes, but sometimes it's interesting to fill with a known
726 pattern for io verification purposes. Depending on the
727 width of the pattern, fio will fill 1/2/3/4 bytes of the
728 buffer at the time. The verify_pattern cannot be larger than
731 verify_fatal=bool Normally fio will keep checking the entire contents
732 before quitting on a block verification failure. If this
733 option is set, fio will exit the job on the first observed
736 stonewall Wait for preceeding jobs in the job file to exit, before
737 starting this one. Can be used to insert serialization
738 points in the job file. A stone wall also implies starting
739 a new reporting group.
741 new_group Start a new reporting group. If this option isn't given,
742 jobs in a file will be part of the same reporting group
743 unless separated by a stone wall (or if it's a group
744 by itself, with the numjobs option).
746 numjobs=int Create the specified number of clones of this job. May be
747 used to setup a larger number of threads/processes doing
748 the same thing. We regard that grouping of jobs as a
751 group_reporting If 'numjobs' is set, it may be interesting to display
752 statistics for the group as a whole instead of for each
753 individual job. This is especially true of 'numjobs' is
754 large, looking at individual thread/process output quickly
755 becomes unwieldy. If 'group_reporting' is specified, fio
756 will show the final report per-group instead of per-job.
758 thread fio defaults to forking jobs, however if this option is
759 given, fio will use pthread_create(3) to create threads
762 zonesize=siint Divide a file into zones of the specified size. See zoneskip.
764 zoneskip=siint Skip the specified number of bytes when zonesize data has
765 been read. The two zone options can be used to only do
766 io on zones of a file.
768 write_iolog=str Write the issued io patterns to the specified file. See
771 read_iolog=str Open an iolog with the specified file name and replay the
772 io patterns it contains. This can be used to store a
773 workload and replay it sometime later. The iolog given
774 may also be a blktrace binary file, which allows fio
775 to replay a workload captured by blktrace. See blktrace
776 for how to capture such logging data. For blktrace replay,
777 the file needs to be turned into a blkparse binary data
778 file first (blktrace <device> -d file_for_fio.bin).
780 write_bw_log If given, write a bandwidth log of the jobs in this job
781 file. Can be used to store data of the bandwidth of the
782 jobs in their lifetime. The included fio_generate_plots
783 script uses gnuplot to turn these text files into nice
786 write_lat_log Same as write_bw_log, except that this option stores io
787 completion latencies instead.
789 lockmem=siint Pin down the specified amount of memory with mlock(2). Can
790 potentially be used instead of removing memory or booting
791 with less memory to simulate a smaller amount of memory.
793 exec_prerun=str Before running this job, issue the command specified
796 exec_postrun=str After the job completes, issue the command specified
799 ioscheduler=str Attempt to switch the device hosting the file to the specified
800 io scheduler before running.
802 cpuload=int If the job is a CPU cycle eater, attempt to use the specified
803 percentage of CPU cycles.
805 cpuchunks=int If the job is a CPU cycle eater, split the load into
806 cycles of the given time. In milliseconds.
808 disk_util=bool Generate disk utilization statistics, if the platform
809 supports it. Defaults to on.
812 6.0 Interpreting the output
813 ---------------------------
815 fio spits out a lot of output. While running, fio will display the
816 status of the jobs created. An example of that would be:
818 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
820 The characters inside the square brackets denote the current status of
821 each thread. The possible values (in typical life cycle order) are:
825 P Thread setup, but not started.
827 I Thread initialized, waiting.
828 R Running, doing sequential reads.
829 r Running, doing random reads.
830 W Running, doing sequential writes.
831 w Running, doing random writes.
832 M Running, doing mixed sequential reads/writes.
833 m Running, doing mixed random reads/writes.
834 F Running, currently waiting for fsync()
835 V Running, doing verification of written data.
836 E Thread exited, not reaped by main thread yet.
839 The other values are fairly self explanatory - number of threads
840 currently running and doing io, rate of io since last check (read speed
841 listed first, then write speed), and the estimated completion percentage
842 and time for the running group. It's impossible to estimate runtime of
843 the following groups (if any).
845 When fio is done (or interrupted by ctrl-c), it will show the data for
846 each thread, group of threads, and disks in that order. For each data
847 direction, the output looks like:
849 Client1 (g=0): err= 0:
850 write: io= 32MiB, bw= 666KiB/s, runt= 50320msec
851 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
852 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
853 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
854 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
855 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
856 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
857 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
858 issued r/w: total=0/32768, short=0/0
859 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
860 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
862 The client number is printed, along with the group id and error of that
863 thread. Below is the io statistics, here for writes. In the order listed,
866 io= Number of megabytes io performed
867 bw= Average bandwidth rate
868 runt= The runtime of that thread
869 slat= Submission latency (avg being the average, stdev being the
870 standard deviation). This is the time it took to submit
871 the io. For sync io, the slat is really the completion
872 latency, since queue/complete is one operation there. This
873 value can be in milliseconds or microseconds, fio will choose
874 the most appropriate base and print that. In the example
875 above, milliseconds is the best scale.
876 clat= Completion latency. Same names as slat, this denotes the
877 time from submission to completion of the io pieces. For
878 sync io, clat will usually be equal (or very close) to 0,
879 as the time from submit to complete is basically just
880 CPU time (io has already been done, see slat explanation).
881 bw= Bandwidth. Same names as the xlat stats, but also includes
882 an approximate percentage of total aggregate bandwidth
883 this thread received in this group. This last value is
884 only really useful if the threads in this group are on the
885 same disk, since they are then competing for disk access.
886 cpu= CPU usage. User and system time, along with the number
887 of context switches this thread went through, usage of
888 system and user time, and finally the number of major
889 and minor page faults.
890 IO depths= The distribution of io depths over the job life time. The
891 numbers are divided into powers of 2, so for example the
892 16= entries includes depths up to that value but higher
893 than the previous entry. In other words, it covers the
895 IO submit= How many pieces of IO were submitting in a single submit
896 call. Each entry denotes that amount and below, until
897 the previous entry - eg, 8=100% mean that we submitted
898 anywhere in between 5-8 ios per submit call.
899 IO complete= Like the above submit number, but for completions instead.
900 IO issued= The number of read/write requests issued, and how many
902 IO latencies= The distribution of IO completion latencies. This is the
903 time from when IO leaves fio and when it gets completed.
904 The numbers follow the same pattern as the IO depths,
905 meaning that 2=1.6% means that 1.6% of the IO completed
906 within 2 msecs, 20=12.8% means that 12.8% of the IO
907 took more than 10 msecs, but less than (or equal to) 20 msecs.
909 After each client has been listed, the group statistics are printed. They
912 Run status group 0 (all jobs):
913 READ: io=64MiB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
914 WRITE: io=64MiB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
916 For each data direction, it prints:
918 io= Number of megabytes io performed.
919 aggrb= Aggregate bandwidth of threads in this group.
920 minb= The minimum average bandwidth a thread saw.
921 maxb= The maximum average bandwidth a thread saw.
922 mint= The smallest runtime of the threads in that group.
923 maxt= The longest runtime of the threads in that group.
925 And finally, the disk statistics are printed. They will look like this:
927 Disk stats (read/write):
928 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
930 Each value is printed for both reads and writes, with reads first. The
933 ios= Number of ios performed by all groups.
934 merge= Number of merges io the io scheduler.
935 ticks= Number of ticks we kept the disk busy.
936 io_queue= Total time spent in the disk queue.
937 util= The disk utilization. A value of 100% means we kept the disk
938 busy constantly, 50% would be a disk idling half of the time.
944 For scripted usage where you typically want to generate tables or graphs
945 of the results, fio can output the results in a semicolon separated format.
946 The format is one long line of values, such as:
948 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%
949 ;0.0%;0.0%;0.0%;0.0%;0.0%
951 To enable terse output, use the --minimal command line option.
953 Split up, the format is as follows:
955 jobname, groupid, error
957 KiB IO, bandwidth (KiB/sec), runtime (msec)
958 Submission latency: min, max, mean, deviation
959 Completion latency: min, max, mean, deviation
960 Bw: min, max, aggregate percentage of total, mean, deviation
962 KiB IO, bandwidth (KiB/sec), runtime (msec)
963 Submission latency: min, max, mean, deviation
964 Completion latency: min, max, mean, deviation
965 Bw: min, max, aggregate percentage of total, mean, deviation
966 CPU usage: user, system, context switches, major faults, minor faults
967 IO depths: <=1, 2, 4, 8, 16, 32, >=64
968 IO latencies: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, >=2000