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 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 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 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 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 int Integer. A whole number value, can be negative. If prefixed with
201 0x, the integer is assumed to be of base 16 (hexadecimal).
202 time Integer with possible time postfix. In seconds unless otherwise
203 specified, use eg 10m for 10 minutes. Accepts s/m/h for seconds,
205 siint SI integer. A whole number value, which may contain a postfix
206 describing the base of the number. Accepted postfixes are k/m/g,
207 meaning kilo, mega, and giga. So if you want to specify 4096,
208 you could either write out '4096' or just give 4k. The postfixes
209 signify base 2 values, so 1024 is 1k and 1024k is 1m and so on.
210 If the option accepts an upper and lower range, use a colon ':'
211 or minus '-' to separate such values. 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 postfix. 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 places 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 and
241 port to connect to in the format of =host/port. If the
242 ioengine is file based, you can specify a number of files
243 by separating the names with a ':' colon. So if you wanted
244 a job to open /dev/sda and /dev/sdb as the two working files,
245 you would use filename=/dev/sda:/dev/sdb. '-' is a reserved
246 name, meaning stdin or stdout. Which of the two depends
247 on the read/write direction set.
249 opendir=str Tell fio to recursively add any file it can find in this
250 directory and down the file system tree.
252 lockfile=str Fio defaults to not doing any locking files before it does
253 IO to them. If a file or file descriptor is shared, fio
254 can serialize IO to that file to make the end result
255 consistent. This is usual for emulating real workloads that
256 share files. The lock modes are:
258 none No locking. The default.
259 exclusive Only one thread/process may do IO,
260 excluding all others.
261 readwrite Read-write locking on the file. Many
262 readers may access the file at the
263 same time, but writes get exclusive
266 The option may be post-fixed with a lock batch number. If
267 set, then each thread/process may do that amount of IOs to
268 the file before giving up the lock. Since lock acquisition is
269 expensive, batching the lock/unlocks will speed up IO.
272 rw=str Type of io pattern. Accepted values are:
274 read Sequential reads
275 write Sequential writes
276 randwrite Random writes
277 randread Random reads
278 rw Sequential mixed reads and writes
279 randrw Random mixed reads and writes
281 For the mixed io types, the default is to split them 50/50.
282 For certain types of io the result may still be skewed a bit,
283 since the speed may be different. It is possible to specify
284 a number of IO's to do before getting a new offset - this
285 is only useful for random IO, where fio would normally
286 generate a new random offset for every IO. If you append
287 eg 8 to randread, you would get a new random offset for
288 every 8 IO's. The result would be a seek for only every 8
289 IO's, instead of for every IO. Use rw=randread:8 to specify
292 randrepeat=bool For random IO workloads, seed the generator in a predictable
293 way so that results are repeatable across repetitions.
295 fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
296 on what IO patterns it is likely to issue. Sometimes you
297 want to test specific IO patterns without telling the
298 kernel about it, in which case you can disable this option.
299 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
300 IO and POSIX_FADV_RANDOM for random IO.
302 size=siint The total size of file io for this job. Fio will run until
303 this many bytes has been transferred, unless runtime is
304 limited by other options (such as 'runtime', for instance).
305 Unless specific nr_files and filesize options are given,
306 fio will divide this size between the available files
307 specified by the job.
309 filesize=siint Individual file sizes. May be a range, in which case fio
310 will select sizes for files at random within the given range
311 and limited to 'size' in total (if that is given). If not
312 given, each created file is the same size.
314 fill_device=bool Sets size to something really large and waits for ENOSPC (no
315 space left on device) as the terminating condition. Only makes
316 sense with sequential write.
319 bs=siint The block size used for the io units. Defaults to 4k. Values
320 can be given for both read and writes. If a single siint is
321 given, it will apply to both. If a second siint is specified
322 after a comma, it will apply to writes only. In other words,
323 the format is either bs=read_and_write or bs=read,write.
324 bs=4k,8k will thus use 4k blocks for reads, and 8k blocks
325 for writes. If you only wish to set the write size, you
326 can do so by passing an empty read size - bs=,8k will set
327 8k for writes and leave the read default value.
329 blocksize_range=irange
330 bsrange=irange Instead of giving a single block size, specify a range
331 and fio will mix the issued io block sizes. The issued
332 io unit will always be a multiple of the minimum value
333 given (also see bs_unaligned). Applies to both reads and
334 writes, however a second range can be given after a comma.
337 bssplit=str Sometimes you want even finer grained control of the
338 block sizes issued, not just an even split between them.
339 This option allows you to weight various block sizes,
340 so that you are able to define a specific amount of
341 block sizes issued. The format for this option is:
343 bssplit=blocksize/percentage:blocksize/percentage
345 for as many block sizes as needed. So if you want to define
346 a workload that has 50% 64k blocks, 10% 4k blocks, and
347 40% 32k blocks, you would write:
349 bssplit=4k/10:64k/50:32k/40
351 Ordering does not matter. If the percentage is left blank,
352 fio will fill in the remaining values evenly. So a bssplit
353 option like this one:
355 bssplit=4k/50:1k/:32k/
357 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
358 always add up to 100, if bssplit is given a range that adds
359 up to more, it will error out.
362 bs_unaligned If this option is given, any byte size value within bsrange
363 may be used as a block range. This typically wont work with
364 direct IO, as that normally requires sector alignment.
366 zero_buffers If this option is given, fio will init the IO buffers to
367 all zeroes. The default is to fill them with random data.
369 refill_buffers If this option is given, fio will refill the IO buffers
370 on every submit. The default is to only fill it at init
371 time and reuse that data. Only makes sense if zero_buffers
372 isn't specified, naturally. If data verification is enabled,
373 refill_buffers is also automatically enabled.
375 nrfiles=int Number of files to use for this job. Defaults to 1.
377 openfiles=int Number of files to keep open at the same time. Defaults to
378 the same as nrfiles, can be set smaller to limit the number
381 file_service_type=str Defines how fio decides which file from a job to
382 service next. The following types are defined:
384 random Just choose a file at random.
386 roundrobin Round robin over open files. This
389 The string can have a number appended, indicating how
390 often to switch to a new file. So if option random:4 is
391 given, fio will switch to a new random file after 4 ios
394 ioengine=str Defines how the job issues io to the file. The following
397 sync Basic read(2) or write(2) io. lseek(2) is
398 used to position the io location.
400 psync Basic pread(2) or pwrite(2) io.
402 vsync Basic readv(2) or writev(2) IO.
404 libaio Linux native asynchronous io.
406 posixaio glibc posix asynchronous io.
408 solarisaio Solaris native asynchronous io.
410 mmap File is memory mapped and data copied
411 to/from using memcpy(3).
413 splice splice(2) is used to transfer the data and
414 vmsplice(2) to transfer data from user
417 syslet-rw Use the syslet system calls to make
418 regular read/write async.
420 sg SCSI generic sg v3 io. May either be
421 synchronous using the SG_IO ioctl, or if
422 the target is an sg character device
423 we use read(2) and write(2) for asynchronous
426 null Doesn't transfer any data, just pretends
427 to. This is mainly used to exercise fio
428 itself and for debugging/testing purposes.
430 net Transfer over the network to given host:port.
431 'filename' must be set appropriately to
432 filename=host/port regardless of send
433 or receive, if the latter only the port
436 netsplice Like net, but uses splice/vmsplice to
437 map data and send/receive.
439 cpuio Doesn't transfer any data, but burns CPU
440 cycles according to the cpuload= and
441 cpucycle= options. Setting cpuload=85
442 will cause that job to do nothing but burn
443 85% of the CPU. In case of SMP machines,
444 use numjobs=<no_of_cpu> to get desired CPU
445 usage, as the cpuload only loads a single
446 CPU at the desired rate.
448 guasi The GUASI IO engine is the Generic Userspace
449 Asyncronous Syscall Interface approach
452 http://www.xmailserver.org/guasi-lib.html
454 for more info on GUASI.
456 external Prefix to specify loading an external
457 IO engine object file. Append the engine
458 filename, eg ioengine=external:/tmp/foo.o
459 to load ioengine foo.o in /tmp.
461 iodepth=int This defines how many io units to keep in flight against
462 the file. The default is 1 for each file defined in this
463 job, can be overridden with a larger value for higher
466 iodepth_batch_submit=int
467 iodepth_batch=int This defines how many pieces of IO to submit at once.
468 It defaults to 1 which means that we submit each IO
469 as soon as it is available, but can be raised to submit
470 bigger batches of IO at the time.
472 iodepth_batch_complete=int This defines how many pieces of IO to retrieve
473 at once. It defaults to 1 which means that we'll ask
474 for a minimum of 1 IO in the retrieval process from
475 the kernel. The IO retrieval will go on until we
476 hit the limit set by iodepth_low. If this variable is
477 set to 0, then fio will always check for completed
478 events before queuing more IO. This helps reduce
479 IO latency, at the cost of more retrieval system calls.
481 iodepth_low=int The low water mark indicating when to start filling
482 the queue again. Defaults to the same as iodepth, meaning
483 that fio will attempt to keep the queue full at all times.
484 If iodepth is set to eg 16 and iodepth_low is set to 4, then
485 after fio has filled the queue of 16 requests, it will let
486 the depth drain down to 4 before starting to fill it again.
488 direct=bool If value is true, use non-buffered io. This is usually
491 buffered=bool If value is true, use buffered io. This is the opposite
492 of the 'direct' option. Defaults to true.
494 offset=siint Start io at the given offset in the file. The data before
495 the given offset will not be touched. This effectively
496 caps the file size at real_size - offset.
498 fsync=int If writing to a file, issue a sync of the dirty data
499 for every number of blocks given. For example, if you give
500 32 as a parameter, fio will sync the file for every 32
501 writes issued. If fio is using non-buffered io, we may
502 not sync the file. The exception is the sg io engine, which
503 synchronizes the disk cache anyway.
505 overwrite=bool If true, writes to a file will always overwrite existing
506 data. If the file doesn't already exist, it will be
507 created before the write phase begins. If the file exists
508 and is large enough for the specified write phase, nothing
511 end_fsync=bool If true, fsync file contents when the job exits.
513 fsync_on_close=bool If true, fio will fsync() a dirty file on close.
514 This differs from end_fsync in that it will happen on every
515 file close, not just at the end of the job.
517 rwmixread=int How large a percentage of the mix should be reads.
519 rwmixwrite=int How large a percentage of the mix should be writes. If both
520 rwmixread and rwmixwrite is given and the values do not add
521 up to 100%, the latter of the two will be used to override
524 norandommap Normally fio will cover every block of the file when doing
525 random IO. If this option is given, fio will just get a
526 new random offset without looking at past io history. This
527 means that some blocks may not be read or written, and that
528 some blocks may be read/written more than once. This option
529 is mutually exclusive with verify= for that reason, since
530 fio doesn't track potential block rewrites which may alter
531 the calculated checksum for that block.
533 softrandommap See norandommap. If fio runs with the random block map enabled
534 and it fails to allocate the map, if this option is set it
535 will continue without a random block map. As coverage will
536 not be as complete as with random maps, this option is
539 nice=int Run the job with the given nice value. See man nice(2).
541 prio=int Set the io priority value of this job. Linux limits us to
542 a positive value between 0 and 7, with 0 being the highest.
545 prioclass=int Set the io priority class. See man ionice(1).
547 thinktime=int Stall the job x microseconds after an io has completed before
548 issuing the next. May be used to simulate processing being
549 done by an application. See thinktime_blocks and
553 Only valid if thinktime is set - pretend to spend CPU time
554 doing something with the data received, before falling back
555 to sleeping for the rest of the period specified by
559 Only valid if thinktime is set - control how many blocks
560 to issue, before waiting 'thinktime' usecs. If not set,
561 defaults to 1 which will make fio wait 'thinktime' usecs
564 rate=int Cap the bandwidth used by this job to this number of KiB/sec.
566 ratemin=int Tell fio to do whatever it can to maintain at least this
567 bandwidth. Failing to meet this requirement, will cause
570 rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
571 as rate, just specified independently of bandwidth. If the
572 job is given a block size range instead of a fixed value,
573 the smallest block size is used as the metric.
575 rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
578 ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
581 cpumask=int Set the CPU affinity of this job. The parameter given is a
582 bitmask of allowed CPU's the job may run on. So if you want
583 the allowed CPUs to be 1 and 5, you would pass the decimal
584 value of (1 << 1 | 1 << 5), or 34. See man
585 sched_setaffinity(2). This may not work on all supported
586 operating systems or kernel versions. This option doesn't
587 work well for a higher CPU count than what you can store in
588 an integer mask, so it can only control cpus 1-32. For
589 boxes with larger CPU counts, use cpus_allowed.
591 cpus_allowed=str Controls the same options as cpumask, but it allows a text
592 setting of the permitted CPUs instead. So to use CPUs 1 and
593 5, you would specify cpus_allowed=1,5. This options also
594 allows a range of CPUs. Say you wanted a binding to CPUs
595 1, 5, and 8-15, you would set cpus_allowed=1,5,8-15.
597 startdelay=time Start this job the specified number of seconds after fio
598 has started. Only useful if the job file contains several
599 jobs, and you want to delay starting some jobs to a certain
602 runtime=time Tell fio to terminate processing after the specified number
603 of seconds. It can be quite hard to determine for how long
604 a specified job will run, so this parameter is handy to
605 cap the total runtime to a given time.
607 time_based If set, fio will run for the duration of the runtime
608 specified even if the file(s) are completely read or
609 written. It will simply loop over the same workload
610 as many times as the runtime allows.
612 ramp_time=time If set, fio will run the specified workload for this amount
613 of time before logging any performance numbers. Useful for
614 letting performance settle before logging results, thus
615 minimizing the runtime required for stable results. Note
616 that the ramp_time is considered lead in time for a job,
617 thus it will increase the total runtime if a special timeout
618 or runtime is specified.
620 invalidate=bool Invalidate the buffer/page cache parts for this file prior
621 to starting io. Defaults to true.
623 sync=bool Use sync io for buffered writes. For the majority of the
624 io engines, this means using O_SYNC.
627 mem=str Fio can use various types of memory as the io unit buffer.
628 The allowed values are:
630 malloc Use memory from malloc(3) as the buffers.
632 shm Use shared memory as the buffers. Allocated
635 shmhuge Same as shm, but use huge pages as backing.
637 mmap Use mmap to allocate buffers. May either be
638 anonymous memory, or can be file backed if
639 a filename is given after the option. The
640 format is mem=mmap:/path/to/file.
642 mmaphuge Use a memory mapped huge file as the buffer
643 backing. Append filename after mmaphuge, ala
644 mem=mmaphuge:/hugetlbfs/file
646 The area allocated is a function of the maximum allowed
647 bs size for the job, multiplied by the io depth given. Note
648 that for shmhuge and mmaphuge to work, the system must have
649 free huge pages allocated. This can normally be checked
650 and set by reading/writing /proc/sys/vm/nr_hugepages on a
651 Linux system. Fio assumes a huge page is 4MiB in size. So
652 to calculate the number of huge pages you need for a given
653 job file, add up the io depth of all jobs (normally one unless
654 iodepth= is used) and multiply by the maximum bs set. Then
655 divide that number by the huge page size. You can see the
656 size of the huge pages in /proc/meminfo. If no huge pages
657 are allocated by having a non-zero number in nr_hugepages,
658 using mmaphuge or shmhuge will fail. Also see hugepage-size.
660 mmaphuge also needs to have hugetlbfs mounted and the file
661 location should point there. So if it's mounted in /huge,
662 you would use mem=mmaphuge:/huge/somefile.
665 Defines the size of a huge page. Must at least be equal
666 to the system setting, see /proc/meminfo. Defaults to 4MiB.
667 Should probably always be a multiple of megabytes, so using
668 hugepage-size=Xm is the preferred way to set this to avoid
669 setting a non-pow-2 bad value.
671 exitall When one job finishes, terminate the rest. The default is
672 to wait for each job to finish, sometimes that is not the
675 bwavgtime=int Average the calculated bandwidth over the given time. Value
676 is specified in milliseconds.
678 create_serialize=bool If true, serialize the file creating for the jobs.
679 This may be handy to avoid interleaving of data
680 files, which may greatly depend on the filesystem
681 used and even the number of processors in the system.
683 create_fsync=bool fsync the data file after creation. This is the
686 unlink=bool Unlink the job files when done. Not the default, as repeated
687 runs of that job would then waste time recreating the file
690 loops=int Run the specified number of iterations of this job. Used
691 to repeat the same workload a given number of times. Defaults
694 do_verify=bool Run the verify phase after a write phase. Only makes sense if
695 verify is set. Defaults to 1.
697 verify=str If writing to a file, fio can verify the file contents
698 after each iteration of the job. The allowed values are:
700 md5 Use an md5 sum of the data area and store
701 it in the header of each block.
703 crc64 Use an experimental crc64 sum of the data
704 area and store it in the header of each
707 crc32c Use a crc32c sum of the data area and store
708 it in the header of each block.
710 crc32c-intel Use hardware assisted crc32c calcuation
711 provided on SSE4.2 enabled processors.
713 crc32 Use a crc32 sum of the data area and store
714 it in the header of each block.
716 crc16 Use a crc16 sum of the data area and store
717 it in the header of each block.
719 crc7 Use a crc7 sum of the data area and store
720 it in the header of each block.
722 sha512 Use sha512 as the checksum function.
724 sha256 Use sha256 as the checksum function.
726 meta Write extra information about each io
727 (timestamp, block number etc.). The block
730 null Only pretend to verify. Useful for testing
731 internals with ioengine=null, not for much
734 This option can be used for repeated burn-in tests of a
735 system to make sure that the written data is also
738 verifysort=bool If set, fio will sort written verify blocks when it deems
739 it faster to read them back in a sorted manner. This is
740 often the case when overwriting an existing file, since
741 the blocks are already laid out in the file system. You
742 can ignore this option unless doing huge amounts of really
743 fast IO where the red-black tree sorting CPU time becomes
746 verify_offset=siint Swap the verification header with data somewhere else
747 in the block before writing. Its swapped back before
750 verify_interval=siint Write the verification header at a finer granularity
751 than the blocksize. It will be written for chunks the
752 size of header_interval. blocksize should divide this
755 verify_pattern=int If set, fio will fill the io buffers with this
756 pattern. Fio defaults to filling with totally random
757 bytes, but sometimes it's interesting to fill with a known
758 pattern for io verification purposes. Depending on the
759 width of the pattern, fio will fill 1/2/3/4 bytes of the
760 buffer at the time. The verify_pattern cannot be larger than
763 verify_fatal=bool Normally fio will keep checking the entire contents
764 before quitting on a block verification failure. If this
765 option is set, fio will exit the job on the first observed
768 stonewall Wait for preceeding jobs in the job file to exit, before
769 starting this one. Can be used to insert serialization
770 points in the job file. A stone wall also implies starting
771 a new reporting group.
773 new_group Start a new reporting group. If this option isn't given,
774 jobs in a file will be part of the same reporting group
775 unless separated by a stone wall (or if it's a group
776 by itself, with the numjobs option).
778 numjobs=int Create the specified number of clones of this job. May be
779 used to setup a larger number of threads/processes doing
780 the same thing. We regard that grouping of jobs as a
783 group_reporting If 'numjobs' is set, it may be interesting to display
784 statistics for the group as a whole instead of for each
785 individual job. This is especially true of 'numjobs' is
786 large, looking at individual thread/process output quickly
787 becomes unwieldy. If 'group_reporting' is specified, fio
788 will show the final report per-group instead of per-job.
790 thread fio defaults to forking jobs, however if this option is
791 given, fio will use pthread_create(3) to create threads
794 zonesize=siint Divide a file into zones of the specified size. See zoneskip.
796 zoneskip=siint Skip the specified number of bytes when zonesize data has
797 been read. The two zone options can be used to only do
798 io on zones of a file.
800 write_iolog=str Write the issued io patterns to the specified file. See
803 read_iolog=str Open an iolog with the specified file name and replay the
804 io patterns it contains. This can be used to store a
805 workload and replay it sometime later. The iolog given
806 may also be a blktrace binary file, which allows fio
807 to replay a workload captured by blktrace. See blktrace
808 for how to capture such logging data. For blktrace replay,
809 the file needs to be turned into a blkparse binary data
810 file first (blktrace <device> -d file_for_fio.bin).
812 write_bw_log=str If given, write a bandwidth log of the jobs in this job
813 file. Can be used to store data of the bandwidth of the
814 jobs in their lifetime. The included fio_generate_plots
815 script uses gnuplot to turn these text files into nice
816 graphs. See write_log_log for behaviour of given
817 filename. For this option, the postfix is _bw.log.
819 write_lat_log=str Same as write_bw_log, except that this option stores io
820 completion latencies instead. If no filename is given
821 with this option, the default filename of "jobname_type.log"
822 is used. Even if the filename is given, fio will still
823 append the type of log. So if one specifies
827 The actual log names will be foo_clat.log and foo_slat.log.
828 This helps fio_generate_plot fine the logs automatically.
830 lockmem=siint Pin down the specified amount of memory with mlock(2). Can
831 potentially be used instead of removing memory or booting
832 with less memory to simulate a smaller amount of memory.
834 exec_prerun=str Before running this job, issue the command specified
837 exec_postrun=str After the job completes, issue the command specified
840 ioscheduler=str Attempt to switch the device hosting the file to the specified
841 io scheduler before running.
843 cpuload=int If the job is a CPU cycle eater, attempt to use the specified
844 percentage of CPU cycles.
846 cpuchunks=int If the job is a CPU cycle eater, split the load into
847 cycles of the given time. In milliseconds.
849 disk_util=bool Generate disk utilization statistics, if the platform
850 supports it. Defaults to on.
852 disable_clat=bool Disable measurements of completion latency numbers. Useful
853 only for cutting back the number of calls to gettimeofday,
854 as that does impact performance at really high IOPS rates.
855 Note that to really get rid of a large amount of these
856 calls, this option must be used with disable_slat and
859 disable_slat=bool Disable measurements of submission latency numbers. See
862 disable_bw=bool Disable measurements of throughput/bandwidth numbers. See
865 gtod_reduce=bool Enable all of the gettimeofday() reducing options
866 (disable_clat, disable_slat, disable_bw) plus reduce
867 precision of the timeout somewhat to really shrink
868 the gettimeofday() call count. With this option enabled,
869 we only do about 0.4% of the gtod() calls we would have
870 done if all time keeping was enabled.
872 gtod_cpu=int Sometimes it's cheaper to dedicate a single thread of
873 execution to just getting the current time. Fio (and
874 databases, for instance) are very intensive on gettimeofday()
875 calls. With this option, you can set one CPU aside for
876 doing nothing but logging current time to a shared memory
877 location. Then the other threads/processes that run IO
878 workloads need only copy that segment, instead of entering
879 the kernel with a gettimeofday() call. The CPU set aside
880 for doing these time calls will be excluded from other
881 uses. Fio will manually clear it from the CPU mask of other
885 6.0 Interpreting the output
886 ---------------------------
888 fio spits out a lot of output. While running, fio will display the
889 status of the jobs created. An example of that would be:
891 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
893 The characters inside the square brackets denote the current status of
894 each thread. The possible values (in typical life cycle order) are:
898 P Thread setup, but not started.
900 I Thread initialized, waiting.
901 R Running, doing sequential reads.
902 r Running, doing random reads.
903 W Running, doing sequential writes.
904 w Running, doing random writes.
905 M Running, doing mixed sequential reads/writes.
906 m Running, doing mixed random reads/writes.
907 F Running, currently waiting for fsync()
908 V Running, doing verification of written data.
909 E Thread exited, not reaped by main thread yet.
912 The other values are fairly self explanatory - number of threads
913 currently running and doing io, rate of io since last check (read speed
914 listed first, then write speed), and the estimated completion percentage
915 and time for the running group. It's impossible to estimate runtime of
916 the following groups (if any).
918 When fio is done (or interrupted by ctrl-c), it will show the data for
919 each thread, group of threads, and disks in that order. For each data
920 direction, the output looks like:
922 Client1 (g=0): err= 0:
923 write: io= 32MiB, bw= 666KiB/s, runt= 50320msec
924 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
925 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
926 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
927 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
928 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
929 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
930 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
931 issued r/w: total=0/32768, short=0/0
932 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
933 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
935 The client number is printed, along with the group id and error of that
936 thread. Below is the io statistics, here for writes. In the order listed,
939 io= Number of megabytes io performed
940 bw= Average bandwidth rate
941 runt= The runtime of that thread
942 slat= Submission latency (avg being the average, stdev being the
943 standard deviation). This is the time it took to submit
944 the io. For sync io, the slat is really the completion
945 latency, since queue/complete is one operation there. This
946 value can be in milliseconds or microseconds, fio will choose
947 the most appropriate base and print that. In the example
948 above, milliseconds is the best scale.
949 clat= Completion latency. Same names as slat, this denotes the
950 time from submission to completion of the io pieces. For
951 sync io, clat will usually be equal (or very close) to 0,
952 as the time from submit to complete is basically just
953 CPU time (io has already been done, see slat explanation).
954 bw= Bandwidth. Same names as the xlat stats, but also includes
955 an approximate percentage of total aggregate bandwidth
956 this thread received in this group. This last value is
957 only really useful if the threads in this group are on the
958 same disk, since they are then competing for disk access.
959 cpu= CPU usage. User and system time, along with the number
960 of context switches this thread went through, usage of
961 system and user time, and finally the number of major
962 and minor page faults.
963 IO depths= The distribution of io depths over the job life time. The
964 numbers are divided into powers of 2, so for example the
965 16= entries includes depths up to that value but higher
966 than the previous entry. In other words, it covers the
968 IO submit= How many pieces of IO were submitting in a single submit
969 call. Each entry denotes that amount and below, until
970 the previous entry - eg, 8=100% mean that we submitted
971 anywhere in between 5-8 ios per submit call.
972 IO complete= Like the above submit number, but for completions instead.
973 IO issued= The number of read/write requests issued, and how many
975 IO latencies= The distribution of IO completion latencies. This is the
976 time from when IO leaves fio and when it gets completed.
977 The numbers follow the same pattern as the IO depths,
978 meaning that 2=1.6% means that 1.6% of the IO completed
979 within 2 msecs, 20=12.8% means that 12.8% of the IO
980 took more than 10 msecs, but less than (or equal to) 20 msecs.
982 After each client has been listed, the group statistics are printed. They
985 Run status group 0 (all jobs):
986 READ: io=64MiB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
987 WRITE: io=64MiB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
989 For each data direction, it prints:
991 io= Number of megabytes io performed.
992 aggrb= Aggregate bandwidth of threads in this group.
993 minb= The minimum average bandwidth a thread saw.
994 maxb= The maximum average bandwidth a thread saw.
995 mint= The smallest runtime of the threads in that group.
996 maxt= The longest runtime of the threads in that group.
998 And finally, the disk statistics are printed. They will look like this:
1000 Disk stats (read/write):
1001 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
1003 Each value is printed for both reads and writes, with reads first. The
1006 ios= Number of ios performed by all groups.
1007 merge= Number of merges io the io scheduler.
1008 ticks= Number of ticks we kept the disk busy.
1009 io_queue= Total time spent in the disk queue.
1010 util= The disk utilization. A value of 100% means we kept the disk
1011 busy constantly, 50% would be a disk idling half of the time.
1017 For scripted usage where you typically want to generate tables or graphs
1018 of the results, fio can output the results in a semicolon separated format.
1019 The format is one long line of values, such as:
1021 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%
1022 ;0.0%;0.0%;0.0%;0.0%;0.0%
1024 To enable terse output, use the --minimal command line option.
1026 Split up, the format is as follows:
1028 jobname, groupid, error
1030 KiB IO, bandwidth (KiB/sec), runtime (msec)
1031 Submission latency: min, max, mean, deviation
1032 Completion latency: min, max, mean, deviation
1033 Bw: min, max, aggregate percentage of total, mean, deviation
1035 KiB IO, bandwidth (KiB/sec), runtime (msec)
1036 Submission latency: min, max, mean, deviation
1037 Completion latency: min, max, mean, deviation
1038 Bw: min, max, aggregate percentage of total, mean, deviation
1039 CPU usage: user, system, context switches, major faults, minor faults
1040 IO depths: <=1, 2, 4, 8, 16, 32, >=64
1041 IO latencies: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, >=2000
projects / fio.git / blob