8 5. Detailed list of parameters
13 1.0 Overview and history
14 ------------------------
15 fio was originally written to save me the hassle of writing special test
16 case programs when I wanted to test a specific workload, either for
17 performance reasons or to find/reproduce a bug. The process of writing
18 such a test app can be tiresome, especially if you have to do it often.
19 Hence I needed a tool that would be able to simulate a given io workload
20 without resorting to writing a tailored test case again and again.
22 A test work load is difficult to define, though. There can be any number
23 of processes or threads involved, and they can each be using their own
24 way of generating io. You could have someone dirtying large amounts of
25 memory in an memory mapped file, or maybe several threads issuing
26 reads using asynchronous io. fio needed to be flexible enough to
27 simulate both of these cases, and many more.
31 The first step in getting fio to simulate a desired io workload, is
32 writing a job file describing that specific setup. A job file may contain
33 any number of threads and/or files - the typical contents of the job file
34 is a global section defining shared parameters, and one or more job
35 sections describing the jobs involved. When run, fio parses this file
36 and sets everything up as described. If we break down a job from top to
37 bottom, it contains the following basic parameters:
39 IO type Defines the io pattern issued to the file(s).
40 We may only be reading sequentially from this
41 file(s), or we may be writing randomly. Or even
42 mixing reads and writes, sequentially or randomly.
44 Block size In how large chunks are we issuing io? This may be
45 a single value, or it may describe a range of
48 IO size How much data are we going to be reading/writing.
50 IO engine How do we issue io? We could be memory mapping the
51 file, we could be using regular read/write, we
52 could be using splice, async io, syslet, or even
55 IO depth If the io engine is async, how large a queuing
56 depth do we want to maintain?
58 IO type Should we be doing buffered io, or direct/raw io?
60 Num files How many files are we spreading the workload over.
62 Num threads How many threads or processes should we spread
65 The above are the basic parameters defined for a workload, in addition
66 there's a multitude of parameters that modify other aspects of how this
72 See the README file for command line parameters, there are only a few
75 Running fio is normally the easiest part - you just give it the job file
76 (or job files) as parameters:
80 and it will start doing what the job_file tells it to do. You can give
81 more than one job file on the command line, fio will serialize the running
82 of those files. Internally that is the same as using the 'stonewall'
83 parameter described the the parameter section.
85 If the job file contains only one job, you may as well just give the
86 parameters on the command line. The command line parameters are identical
87 to the job parameters, with a few extra that control global parameters
88 (see README). For example, for the job file parameter iodepth=2, the
89 mirror command line option would be --iodepth 2 or --iodepth=2. You can
90 also use the command line for giving more than one job entry. For each
91 --name option that fio sees, it will start a new job with that name.
92 Command line entries following a --name entry will apply to that job,
93 until there are no more entries or a new --name entry is seen. This is
94 similar to the job file options, where each option applies to the current
95 job until a new [] job entry is seen.
97 fio does not need to run as root, except if the files or devices specified
98 in the job section requires that. Some other options may also be restricted,
99 such as memory locking, io scheduler switching, and decreasing the nice value.
104 As previously described, fio accepts one or more job files describing
105 what it is supposed to do. The job file format is the classic ini file,
106 where the names enclosed in [] brackets define the job name. You are free
107 to use any ascii name you want, except 'global' which has special meaning.
108 A global section sets defaults for the jobs described in that file. A job
109 may override a global section parameter, and a job file may even have
110 several global sections if so desired. A job is only affected by a global
111 section residing above it. If the first character in a line is a ';' or a
112 '#', the entire line is discarded as a comment.
114 So let's look at a really simple job file that defines two processes, each
115 randomly reading from a 128MB file.
117 ; -- start job file --
128 As you can see, the job file sections themselves are empty as all the
129 described parameters are shared. As no filename= option is given, fio
130 makes up a filename for each of the jobs as it sees fit. On the command
131 line, this job would look as follows:
133 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
136 Let's look at an example that has a number of processes writing randomly
139 ; -- start job file --
151 Here we have no global section, as we only have one job defined anyway.
152 We want to use async io here, with a depth of 4 for each file. We also
153 increased the buffer size used to 32KB and define numjobs to 4 to
154 fork 4 identical jobs. The result is 4 processes each randomly writing
155 to their own 64MB file. Instead of using the above job file, you could
156 have given the parameters on the command line. For this case, you would
159 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
161 4.1 Environment variables
162 -------------------------
164 fio also supports environment variable expansion in job files. Any
165 substring of the form "${VARNAME}" as part of an option value (in other
166 words, on the right of the `='), will be expanded to the value of the
167 environment variable called VARNAME. If no such environment variable
168 is defined, or VARNAME is the empty string, the empty string will be
171 As an example, let's look at a sample fio invocation and job file:
173 $ SIZE=64m NUMJOBS=4 fio jobfile.fio
175 ; -- start job file --
182 This will expand to the following equivalent job file at runtime:
184 ; -- start job file --
191 fio ships with a few example job files, you can also look there for
194 4.2 Reserved keywords
195 ---------------------
197 Additionally, fio has a set of reserved keywords that will be replaced
198 internally with the appropriate value. Those keywords are:
200 $pagesize The architecture page size of the running system
201 $mb_memory Megabytes of total memory in the system
202 $ncpus Number of online available CPUs
204 These can be used on the command line or in the job file, and will be
205 automatically substituted with the current system values when the job
206 is run. Simple math is also supported on these keywords, so you can
207 perform actions like:
211 and get that properly expanded to 8 times the size of memory in the
215 5.0 Detailed list of parameters
216 -------------------------------
218 This section describes in details each parameter associated with a job.
219 Some parameters take an option of a given type, such as an integer or
220 a string. The following types are used:
222 str String. This is a sequence of alpha characters.
223 time Integer with possible time suffix. In seconds unless otherwise
224 specified, use eg 10m for 10 minutes. Accepts s/m/h for seconds,
226 int SI integer. A whole number value, which may contain a suffix
227 describing the base of the number. Accepted suffixes are k/m/g/t/p,
228 meaning kilo, mega, giga, tera, and peta. The suffix is not case
229 sensitive, and you may also include trailing 'b' (eg 'kb' is the same
230 as 'k'). So if you want to specify 4096, you could either write
231 out '4096' or just give 4k. The suffixes signify base 2 values, so
232 1024 is 1k and 1024k is 1m and so on, unless the suffix is explicitly
233 set to a base 10 value using 'kib', 'mib', 'gib', etc. If that is the
234 case, then 1000 is used as the multiplier. This can be handy for
235 disks, since manufacturers generally use base 10 values when listing
236 the capacity of a drive. If the option accepts an upper and lower
237 range, use a colon ':' or minus '-' to separate such values. May also
238 include a prefix to indicate numbers base. If 0x is used, the number
239 is assumed to be hexadecimal. See irange.
240 bool Boolean. Usually parsed as an integer, however only defined for
241 true and false (1 and 0).
242 irange Integer range with suffix. Allows value range to be given, such
243 as 1024-4096. A colon may also be used as the separator, eg
244 1k:4k. If the option allows two sets of ranges, they can be
245 specified with a ',' or '/' delimiter: 1k-4k/8k-32k. Also see
248 With the above in mind, here follows the complete list of fio job
251 name=str ASCII name of the job. This may be used to override the
252 name printed by fio for this job. Otherwise the job
253 name is used. On the command line this parameter has the
254 special purpose of also signaling the start of a new
257 description=str Text description of the job. Doesn't do anything except
258 dump this text description when this job is run. It's
261 directory=str Prefix filenames with this directory. Used to place files
262 in a different location than "./".
264 filename=str Fio normally makes up a filename based on the job name,
265 thread number, and file number. If you want to share
266 files between threads in a job or several jobs, specify
267 a filename for each of them to override the default. If
268 the ioengine used is 'net', the filename is the host, port,
269 and protocol to use in the format of =host/port/protocol.
270 See ioengine=net for more. If the ioengine is file based, you
271 can specify a number of files by separating the names with a
272 ':' colon. So if you wanted a job to open /dev/sda and /dev/sdb
273 as the two working files, you would use
274 filename=/dev/sda:/dev/sdb. If the wanted filename does need to
275 include a colon, then escape that with a '\' character. For
276 instance, if the filename is "/dev/dsk/foo@3,0:c", then you would
277 use filename="/dev/dsk/foo@3,0\:c". '-' is a reserved name,
278 meaning stdin or stdout. Which of the two depends on the read/write
281 opendir=str Tell fio to recursively add any file it can find in this
282 directory and down the file system tree.
284 lockfile=str Fio defaults to not locking any files before it does
285 IO to them. If a file or file descriptor is shared, fio
286 can serialize IO to that file to make the end result
287 consistent. This is usual for emulating real workloads that
288 share files. The lock modes are:
290 none No locking. The default.
291 exclusive Only one thread/process may do IO,
292 excluding all others.
293 readwrite Read-write locking on the file. Many
294 readers may access the file at the
295 same time, but writes get exclusive
298 The option may be post-fixed with a lock batch number. If
299 set, then each thread/process may do that amount of IOs to
300 the file before giving up the lock. Since lock acquisition is
301 expensive, batching the lock/unlocks will speed up IO.
304 rw=str Type of io pattern. Accepted values are:
306 read Sequential reads
307 write Sequential writes
308 randwrite Random writes
309 randread Random reads
310 rw Sequential mixed reads and writes
311 randrw Random mixed reads and writes
313 For the mixed io types, the default is to split them 50/50.
314 For certain types of io the result may still be skewed a bit,
315 since the speed may be different. It is possible to specify
316 a number of IO's to do before getting a new offset - this
317 is only useful for random IO, where fio would normally
318 generate a new random offset for every IO. If you append
319 eg 8 to randread, you would get a new random offset for
320 every 8 IO's. The result would be a seek for only every 8
321 IO's, instead of for every IO. Use rw=randread:8 to specify
324 kb_base=int The base unit for a kilobyte. The defacto base is 2^10, 1024.
325 Storage manufacturers like to use 10^3 or 1000 as a base
326 ten unit instead, for obvious reasons. Allow values are
327 1024 or 1000, with 1024 being the default.
329 randrepeat=bool For random IO workloads, seed the generator in a predictable
330 way so that results are repeatable across repetitions.
332 fallocate=bool By default, fio will use fallocate() to advise the system
333 of the size of the file we are going to write. This can be
334 turned off with fallocate=0. May not be available on all
337 fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
338 on what IO patterns it is likely to issue. Sometimes you
339 want to test specific IO patterns without telling the
340 kernel about it, in which case you can disable this option.
341 If set, fio will use POSIX_FADV_SEQUENTIAL for sequential
342 IO and POSIX_FADV_RANDOM for random IO.
344 size=int The total size of file io for this job. Fio will run until
345 this many bytes has been transferred, unless runtime is
346 limited by other options (such as 'runtime', for instance).
347 Unless specific nrfiles and filesize options are given,
348 fio will divide this size between the available files
349 specified by the job. If not set, fio will use the full
350 size of the given files or devices. If the the files
351 do not exist, size must be given.
353 filesize=int Individual file sizes. May be a range, in which case fio
354 will select sizes for files at random within the given range
355 and limited to 'size' in total (if that is given). If not
356 given, each created file is the same size.
358 fill_device=bool Sets size to something really large and waits for ENOSPC (no
359 space left on device) as the terminating condition. Only makes
360 sense with sequential write. For a read workload, the mount
361 point will be filled first then IO started on the result.
364 bs=int The block size used for the io units. Defaults to 4k. Values
365 can be given for both read and writes. If a single int is
366 given, it will apply to both. If a second int is specified
367 after a comma, it will apply to writes only. In other words,
368 the format is either bs=read_and_write or bs=read,write.
369 bs=4k,8k will thus use 4k blocks for reads, and 8k blocks
370 for writes. If you only wish to set the write size, you
371 can do so by passing an empty read size - bs=,8k will set
372 8k for writes and leave the read default value.
375 ba=int At what boundary to align random IO offsets. Defaults to
376 the same as 'blocksize' the minimum blocksize given.
377 Minimum alignment is typically 512b for using direct IO,
378 though it usually depends on the hardware block size. This
379 option is mutually exclusive with using a random map for
380 files, so it will turn off that option.
382 blocksize_range=irange
383 bsrange=irange Instead of giving a single block size, specify a range
384 and fio will mix the issued io block sizes. The issued
385 io unit will always be a multiple of the minimum value
386 given (also see bs_unaligned). Applies to both reads and
387 writes, however a second range can be given after a comma.
390 bssplit=str Sometimes you want even finer grained control of the
391 block sizes issued, not just an even split between them.
392 This option allows you to weight various block sizes,
393 so that you are able to define a specific amount of
394 block sizes issued. The format for this option is:
396 bssplit=blocksize/percentage:blocksize/percentage
398 for as many block sizes as needed. So if you want to define
399 a workload that has 50% 64k blocks, 10% 4k blocks, and
400 40% 32k blocks, you would write:
402 bssplit=4k/10:64k/50:32k/40
404 Ordering does not matter. If the percentage is left blank,
405 fio will fill in the remaining values evenly. So a bssplit
406 option like this one:
408 bssplit=4k/50:1k/:32k/
410 would have 50% 4k ios, and 25% 1k and 32k ios. The percentages
411 always add up to 100, if bssplit is given a range that adds
412 up to more, it will error out.
414 bssplit also supports giving separate splits to reads and
415 writes. The format is identical to what bs= accepts. You
416 have to separate the read and write parts with a comma. So
417 if you want a workload that has 50% 2k reads and 50% 4k reads,
418 while having 90% 4k writes and 10% 8k writes, you would
421 bssplit=2k/50:4k/50,4k/90,8k/10
424 bs_unaligned If this option is given, any byte size value within bsrange
425 may be used as a block range. This typically wont work with
426 direct IO, as that normally requires sector alignment.
428 zero_buffers If this option is given, fio will init the IO buffers to
429 all zeroes. The default is to fill them with random data.
431 refill_buffers If this option is given, fio will refill the IO buffers
432 on every submit. The default is to only fill it at init
433 time and reuse that data. Only makes sense if zero_buffers
434 isn't specified, naturally. If data verification is enabled,
435 refill_buffers is also automatically enabled.
437 nrfiles=int Number of files to use for this job. Defaults to 1.
439 openfiles=int Number of files to keep open at the same time. Defaults to
440 the same as nrfiles, can be set smaller to limit the number
443 file_service_type=str Defines how fio decides which file from a job to
444 service next. The following types are defined:
446 random Just choose a file at random.
448 roundrobin Round robin over open files. This
451 sequential Finish one file before moving on to
452 the next. Multiple files can still be
453 open depending on 'openfiles'.
455 The string can have a number appended, indicating how
456 often to switch to a new file. So if option random:4 is
457 given, fio will switch to a new random file after 4 ios
460 ioengine=str Defines how the job issues io to the file. The following
463 sync Basic read(2) or write(2) io. lseek(2) is
464 used to position the io location.
466 psync Basic pread(2) or pwrite(2) io.
468 vsync Basic readv(2) or writev(2) IO.
470 libaio Linux native asynchronous io. Note that Linux
471 may only support queued behaviour with
472 non-buffered IO (set direct=1 or buffered=0).
474 posixaio glibc posix asynchronous io.
476 solarisaio Solaris native asynchronous io.
478 mmap File is memory mapped and data copied
479 to/from using memcpy(3).
481 splice splice(2) is used to transfer the data and
482 vmsplice(2) to transfer data from user
485 syslet-rw Use the syslet system calls to make
486 regular read/write async.
488 sg SCSI generic sg v3 io. May either be
489 synchronous using the SG_IO ioctl, or if
490 the target is an sg character device
491 we use read(2) and write(2) for asynchronous
494 null Doesn't transfer any data, just pretends
495 to. This is mainly used to exercise fio
496 itself and for debugging/testing purposes.
498 net Transfer over the network to given host:port.
499 'filename' must be set appropriately to
500 filename=host/port/protocol regardless of send
501 or receive, if the latter only the port
502 argument is used. 'host' may be an IP address
503 or hostname, port is the port number to be used,
504 and protocol may be 'udp' or 'tcp'. If no
505 protocol is given, TCP is used.
507 netsplice Like net, but uses splice/vmsplice to
508 map data and send/receive.
510 cpuio Doesn't transfer any data, but burns CPU
511 cycles according to the cpuload= and
512 cpucycle= options. Setting cpuload=85
513 will cause that job to do nothing but burn
514 85% of the CPU. In case of SMP machines,
515 use numjobs=<no_of_cpu> to get desired CPU
516 usage, as the cpuload only loads a single
517 CPU at the desired rate.
519 guasi The GUASI IO engine is the Generic Userspace
520 Asyncronous Syscall Interface approach
523 http://www.xmailserver.org/guasi-lib.html
525 for more info on GUASI.
527 external Prefix to specify loading an external
528 IO engine object file. Append the engine
529 filename, eg ioengine=external:/tmp/foo.o
530 to load ioengine foo.o in /tmp.
532 iodepth=int This defines how many io units to keep in flight against
533 the file. The default is 1 for each file defined in this
534 job, can be overridden with a larger value for higher
537 iodepth_batch_submit=int
538 iodepth_batch=int This defines how many pieces of IO to submit at once.
539 It defaults to 1 which means that we submit each IO
540 as soon as it is available, but can be raised to submit
541 bigger batches of IO at the time.
543 iodepth_batch_complete=int This defines how many pieces of IO to retrieve
544 at once. It defaults to 1 which means that we'll ask
545 for a minimum of 1 IO in the retrieval process from
546 the kernel. The IO retrieval will go on until we
547 hit the limit set by iodepth_low. If this variable is
548 set to 0, then fio will always check for completed
549 events before queuing more IO. This helps reduce
550 IO latency, at the cost of more retrieval system calls.
552 iodepth_low=int The low water mark indicating when to start filling
553 the queue again. Defaults to the same as iodepth, meaning
554 that fio will attempt to keep the queue full at all times.
555 If iodepth is set to eg 16 and iodepth_low is set to 4, then
556 after fio has filled the queue of 16 requests, it will let
557 the depth drain down to 4 before starting to fill it again.
559 direct=bool If value is true, use non-buffered io. This is usually
562 buffered=bool If value is true, use buffered io. This is the opposite
563 of the 'direct' option. Defaults to true.
565 offset=int Start io at the given offset in the file. The data before
566 the given offset will not be touched. This effectively
567 caps the file size at real_size - offset.
569 fsync=int If writing to a file, issue a sync of the dirty data
570 for every number of blocks given. For example, if you give
571 32 as a parameter, fio will sync the file for every 32
572 writes issued. If fio is using non-buffered io, we may
573 not sync the file. The exception is the sg io engine, which
574 synchronizes the disk cache anyway.
576 fdatasync=int Like fsync= but uses fdatasync() to only sync data and not
578 In FreeBSD there is no fdatasync(), this falls back to
581 sync_file_range=str:val Use sync_file_range() for every 'val' number of
582 write operations. Fio will track range of writes that
583 have happened since the last sync_file_range() call. 'str'
584 can currently be one or more of:
586 wait_before SYNC_FILE_RANGE_WAIT_BEFORE
587 write SYNC_FILE_RANGE_WRITE
588 wait_after SYNC_FILE_RANGE_WAIT_AFTER
590 So if you do sync_file_range=wait_before,write:8, fio would
591 use SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE for
592 every 8 writes. Also see the sync_file_range(2) man page.
593 This option is Linux specific.
595 overwrite=bool If true, writes to a file will always overwrite existing
596 data. If the file doesn't already exist, it will be
597 created before the write phase begins. If the file exists
598 and is large enough for the specified write phase, nothing
601 end_fsync=bool If true, fsync file contents when the job exits.
603 fsync_on_close=bool If true, fio will fsync() a dirty file on close.
604 This differs from end_fsync in that it will happen on every
605 file close, not just at the end of the job.
607 rwmixread=int How large a percentage of the mix should be reads.
609 rwmixwrite=int How large a percentage of the mix should be writes. If both
610 rwmixread and rwmixwrite is given and the values do not add
611 up to 100%, the latter of the two will be used to override
612 the first. This may interfere with a given rate setting,
613 if fio is asked to limit reads or writes to a certain rate.
614 If that is the case, then the distribution may be skewed.
616 norandommap Normally fio will cover every block of the file when doing
617 random IO. If this option is given, fio will just get a
618 new random offset without looking at past io history. This
619 means that some blocks may not be read or written, and that
620 some blocks may be read/written more than once. This option
621 is mutually exclusive with verify= if and only if multiple
622 blocksizes (via bsrange=) are used, since fio only tracks
623 complete rewrites of blocks.
625 softrandommap See norandommap. If fio runs with the random block map enabled
626 and it fails to allocate the map, if this option is set it
627 will continue without a random block map. As coverage will
628 not be as complete as with random maps, this option is
631 nice=int Run the job with the given nice value. See man nice(2).
633 prio=int Set the io priority value of this job. Linux limits us to
634 a positive value between 0 and 7, with 0 being the highest.
637 prioclass=int Set the io priority class. See man ionice(1).
639 thinktime=int Stall the job x microseconds after an io has completed before
640 issuing the next. May be used to simulate processing being
641 done by an application. See thinktime_blocks and
645 Only valid if thinktime is set - pretend to spend CPU time
646 doing something with the data received, before falling back
647 to sleeping for the rest of the period specified by
651 Only valid if thinktime is set - control how many blocks
652 to issue, before waiting 'thinktime' usecs. If not set,
653 defaults to 1 which will make fio wait 'thinktime' usecs
656 rate=int Cap the bandwidth used by this job. The number is in bytes/sec,
657 the normal suffix rules apply. You can use rate=500k to limit
658 reads and writes to 500k each, or you can specify read and
659 writes separately. Using rate=1m,500k would limit reads to
660 1MB/sec and writes to 500KB/sec. Capping only reads or
661 writes can be done with rate=,500k or rate=500k,. The former
662 will only limit writes (to 500KB/sec), the latter will only
665 ratemin=int Tell fio to do whatever it can to maintain at least this
666 bandwidth. Failing to meet this requirement, will cause
667 the job to exit. The same format as rate is used for
668 read vs write separation.
670 rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same
671 as rate, just specified independently of bandwidth. If the
672 job is given a block size range instead of a fixed value,
673 the smallest block size is used as the metric. The same format
674 as rate is used for read vs write seperation.
676 rate_iops_min=int If fio doesn't meet this rate of IO, it will cause
677 the job to exit. The same format as rate is used for read vs
680 ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
683 cpumask=int Set the CPU affinity of this job. The parameter given is a
684 bitmask of allowed CPU's the job may run on. So if you want
685 the allowed CPUs to be 1 and 5, you would pass the decimal
686 value of (1 << 1 | 1 << 5), or 34. See man
687 sched_setaffinity(2). This may not work on all supported
688 operating systems or kernel versions. This option doesn't
689 work well for a higher CPU count than what you can store in
690 an integer mask, so it can only control cpus 1-32. For
691 boxes with larger CPU counts, use cpus_allowed.
693 cpus_allowed=str Controls the same options as cpumask, but it allows a text
694 setting of the permitted CPUs instead. So to use CPUs 1 and
695 5, you would specify cpus_allowed=1,5. This options also
696 allows a range of CPUs. Say you wanted a binding to CPUs
697 1, 5, and 8-15, you would set cpus_allowed=1,5,8-15.
699 startdelay=time Start this job the specified number of seconds after fio
700 has started. Only useful if the job file contains several
701 jobs, and you want to delay starting some jobs to a certain
704 runtime=time Tell fio to terminate processing after the specified number
705 of seconds. It can be quite hard to determine for how long
706 a specified job will run, so this parameter is handy to
707 cap the total runtime to a given time.
709 time_based If set, fio will run for the duration of the runtime
710 specified even if the file(s) are completely read or
711 written. It will simply loop over the same workload
712 as many times as the runtime allows.
714 ramp_time=time If set, fio will run the specified workload for this amount
715 of time before logging any performance numbers. Useful for
716 letting performance settle before logging results, thus
717 minimizing the runtime required for stable results. Note
718 that the ramp_time is considered lead in time for a job,
719 thus it will increase the total runtime if a special timeout
720 or runtime is specified.
722 invalidate=bool Invalidate the buffer/page cache parts for this file prior
723 to starting io. Defaults to true.
725 sync=bool Use sync io for buffered writes. For the majority of the
726 io engines, this means using O_SYNC.
729 mem=str Fio can use various types of memory as the io unit buffer.
730 The allowed values are:
732 malloc Use memory from malloc(3) as the buffers.
734 shm Use shared memory as the buffers. Allocated
737 shmhuge Same as shm, but use huge pages as backing.
739 mmap Use mmap to allocate buffers. May either be
740 anonymous memory, or can be file backed if
741 a filename is given after the option. The
742 format is mem=mmap:/path/to/file.
744 mmaphuge Use a memory mapped huge file as the buffer
745 backing. Append filename after mmaphuge, ala
746 mem=mmaphuge:/hugetlbfs/file
748 The area allocated is a function of the maximum allowed
749 bs size for the job, multiplied by the io depth given. Note
750 that for shmhuge and mmaphuge to work, the system must have
751 free huge pages allocated. This can normally be checked
752 and set by reading/writing /proc/sys/vm/nr_hugepages on a
753 Linux system. Fio assumes a huge page is 4MB in size. So
754 to calculate the number of huge pages you need for a given
755 job file, add up the io depth of all jobs (normally one unless
756 iodepth= is used) and multiply by the maximum bs set. Then
757 divide that number by the huge page size. You can see the
758 size of the huge pages in /proc/meminfo. If no huge pages
759 are allocated by having a non-zero number in nr_hugepages,
760 using mmaphuge or shmhuge will fail. Also see hugepage-size.
762 mmaphuge also needs to have hugetlbfs mounted and the file
763 location should point there. So if it's mounted in /huge,
764 you would use mem=mmaphuge:/huge/somefile.
766 iomem_align=int This indiciates the memory alignment of the IO memory buffers.
767 Note that the given alignment is applied to the first IO unit
768 buffer, if using iodepth the alignment of the following buffers
769 are given by the bs used. In other words, if using a bs that is
770 a multiple of the page sized in the system, all buffers will
771 be aligned to this value. If using a bs that is not page
772 aligned, the alignment of subsequent IO memory buffers is the
773 sum of the iomem_align and bs used.
776 Defines the size of a huge page. Must at least be equal
777 to the system setting, see /proc/meminfo. Defaults to 4MB.
778 Should probably always be a multiple of megabytes, so using
779 hugepage-size=Xm is the preferred way to set this to avoid
780 setting a non-pow-2 bad value.
782 exitall When one job finishes, terminate the rest. The default is
783 to wait for each job to finish, sometimes that is not the
786 bwavgtime=int Average the calculated bandwidth over the given time. Value
787 is specified in milliseconds.
789 create_serialize=bool If true, serialize the file creating for the jobs.
790 This may be handy to avoid interleaving of data
791 files, which may greatly depend on the filesystem
792 used and even the number of processors in the system.
794 create_fsync=bool fsync the data file after creation. This is the
797 create_on_open=bool Don't pre-setup the files for IO, just create open()
798 when it's time to do IO to that file.
800 pre_read=bool If this is given, files will be pre-read into memory before
801 starting the given IO operation. This will also clear
802 the 'invalidate' flag, since it is pointless to pre-read
803 and then drop the cache. This will only work for IO engines
804 that are seekable, since they allow you to read the same data
805 multiple times. Thus it will not work on eg network or splice
808 unlink=bool Unlink the job files when done. Not the default, as repeated
809 runs of that job would then waste time recreating the file
812 loops=int Run the specified number of iterations of this job. Used
813 to repeat the same workload a given number of times. Defaults
816 do_verify=bool Run the verify phase after a write phase. Only makes sense if
817 verify is set. Defaults to 1.
819 verify=str If writing to a file, fio can verify the file contents
820 after each iteration of the job. The allowed values are:
822 md5 Use an md5 sum of the data area and store
823 it in the header of each block.
825 crc64 Use an experimental crc64 sum of the data
826 area and store it in the header of each
829 crc32c Use a crc32c sum of the data area and store
830 it in the header of each block.
832 crc32c-intel Use hardware assisted crc32c calcuation
833 provided on SSE4.2 enabled processors. Falls
834 back to regular software crc32c, if not
835 supported by the system.
837 crc32 Use a crc32 sum of the data area and store
838 it in the header of each block.
840 crc16 Use a crc16 sum of the data area and store
841 it in the header of each block.
843 crc7 Use a crc7 sum of the data area and store
844 it in the header of each block.
846 sha512 Use sha512 as the checksum function.
848 sha256 Use sha256 as the checksum function.
850 sha1 Use optimized sha1 as the checksum function.
852 meta Write extra information about each io
853 (timestamp, block number etc.). The block
854 number is verified. See also verify_pattern.
856 null Only pretend to verify. Useful for testing
857 internals with ioengine=null, not for much
860 This option can be used for repeated burn-in tests of a
861 system to make sure that the written data is also
862 correctly read back. If the data direction given is
863 a read or random read, fio will assume that it should
864 verify a previously written file. If the data direction
865 includes any form of write, the verify will be of the
868 verifysort=bool If set, fio will sort written verify blocks when it deems
869 it faster to read them back in a sorted manner. This is
870 often the case when overwriting an existing file, since
871 the blocks are already laid out in the file system. You
872 can ignore this option unless doing huge amounts of really
873 fast IO where the red-black tree sorting CPU time becomes
876 verify_offset=int Swap the verification header with data somewhere else
877 in the block before writing. Its swapped back before
880 verify_interval=int Write the verification header at a finer granularity
881 than the blocksize. It will be written for chunks the
882 size of header_interval. blocksize should divide this
885 verify_pattern=str If set, fio will fill the io buffers with this
886 pattern. Fio defaults to filling with totally random
887 bytes, but sometimes it's interesting to fill with a known
888 pattern for io verification purposes. Depending on the
889 width of the pattern, fio will fill 1/2/3/4 bytes of the
890 buffer at the time(it can be either a decimal or a hex number).
891 The verify_pattern if larger than a 32-bit quantity has to
892 be a hex number that starts with either "0x" or "0X". Use
895 verify_fatal=bool Normally fio will keep checking the entire contents
896 before quitting on a block verification failure. If this
897 option is set, fio will exit the job on the first observed
900 verify_async=int Fio will normally verify IO inline from the submitting
901 thread. This option takes an integer describing how many
902 async offload threads to create for IO verification instead,
903 causing fio to offload the duty of verifying IO contents
904 to one or more separate threads. If using this offload
905 option, even sync IO engines can benefit from using an
906 iodepth setting higher than 1, as it allows them to have
907 IO in flight while verifies are running.
909 verify_async_cpus=str Tell fio to set the given CPU affinity on the
910 async IO verification threads. See cpus_allowed for the
913 verify_backlog=int Fio will normally verify the written contents of a
914 job that utilizes verify once that job has completed. In
915 other words, everything is written then everything is read
916 back and verified. You may want to verify continually
917 instead for a variety of reasons. Fio stores the meta data
918 associated with an IO block in memory, so for large
919 verify workloads, quite a bit of memory would be used up
920 holding this meta data. If this option is enabled, fio
921 will verify the previously written blocks before continuing
924 verify_backlog_batch=int Control how many blocks fio will verify
925 if verify_backlog is set. If not set, will default to
926 the value of verify_backlog (meaning the entire queue
927 is read back and verified).
929 stonewall Wait for preceeding jobs in the job file to exit, before
930 starting this one. Can be used to insert serialization
931 points in the job file. A stone wall also implies starting
932 a new reporting group.
934 new_group Start a new reporting group. If this option isn't given,
935 jobs in a file will be part of the same reporting group
936 unless separated by a stone wall (or if it's a group
937 by itself, with the numjobs option).
939 numjobs=int Create the specified number of clones of this job. May be
940 used to setup a larger number of threads/processes doing
941 the same thing. We regard that grouping of jobs as a
944 group_reporting If 'numjobs' is set, it may be interesting to display
945 statistics for the group as a whole instead of for each
946 individual job. This is especially true of 'numjobs' is
947 large, looking at individual thread/process output quickly
948 becomes unwieldy. If 'group_reporting' is specified, fio
949 will show the final report per-group instead of per-job.
951 thread fio defaults to forking jobs, however if this option is
952 given, fio will use pthread_create(3) to create threads
955 zonesize=int Divide a file into zones of the specified size. See zoneskip.
957 zoneskip=int Skip the specified number of bytes when zonesize data has
958 been read. The two zone options can be used to only do
959 io on zones of a file.
961 write_iolog=str Write the issued io patterns to the specified file. See
964 read_iolog=str Open an iolog with the specified file name and replay the
965 io patterns it contains. This can be used to store a
966 workload and replay it sometime later. The iolog given
967 may also be a blktrace binary file, which allows fio
968 to replay a workload captured by blktrace. See blktrace
969 for how to capture such logging data. For blktrace replay,
970 the file needs to be turned into a blkparse binary data
971 file first (blkparse <device> -o /dev/null -d file_for_fio.bin).
973 write_bw_log=str If given, write a bandwidth log of the jobs in this job
974 file. Can be used to store data of the bandwidth of the
975 jobs in their lifetime. The included fio_generate_plots
976 script uses gnuplot to turn these text files into nice
977 graphs. See write_log_log for behaviour of given
978 filename. For this option, the postfix is _bw.log.
980 write_lat_log=str Same as write_bw_log, except that this option stores io
981 submission, completion, and total latencies instead. If no
982 filename is given with this option, the default filename of
983 "jobname_type.log" is used. Even if the filename is given,
984 fio will still append the type of log. So if one specifies
988 The actual log names will be foo_slat.log, foo_slat.log,
989 and foo_lat.log. This helps fio_generate_plot fine the logs
992 lockmem=int Pin down the specified amount of memory with mlock(2). Can
993 potentially be used instead of removing memory or booting
994 with less memory to simulate a smaller amount of memory.
996 exec_prerun=str Before running this job, issue the command specified
999 exec_postrun=str After the job completes, issue the command specified
1002 ioscheduler=str Attempt to switch the device hosting the file to the specified
1003 io scheduler before running.
1005 cpuload=int If the job is a CPU cycle eater, attempt to use the specified
1006 percentage of CPU cycles.
1008 cpuchunks=int If the job is a CPU cycle eater, split the load into
1009 cycles of the given time. In microseconds.
1011 disk_util=bool Generate disk utilization statistics, if the platform
1012 supports it. Defaults to on.
1014 disable_lat=bool Disable measurements of total latency numbers. Useful
1015 only for cutting back the number of calls to gettimeofday,
1016 as that does impact performance at really high IOPS rates.
1017 Note that to really get rid of a large amount of these
1018 calls, this option must be used with disable_slat and
1021 disable_clat=bool Disable measurements of completion latency numbers. See
1024 disable_slat=bool Disable measurements of submission latency numbers. See
1027 disable_bw=bool Disable measurements of throughput/bandwidth numbers. See
1030 gtod_reduce=bool Enable all of the gettimeofday() reducing options
1031 (disable_clat, disable_slat, disable_bw) plus reduce
1032 precision of the timeout somewhat to really shrink
1033 the gettimeofday() call count. With this option enabled,
1034 we only do about 0.4% of the gtod() calls we would have
1035 done if all time keeping was enabled.
1037 gtod_cpu=int Sometimes it's cheaper to dedicate a single thread of
1038 execution to just getting the current time. Fio (and
1039 databases, for instance) are very intensive on gettimeofday()
1040 calls. With this option, you can set one CPU aside for
1041 doing nothing but logging current time to a shared memory
1042 location. Then the other threads/processes that run IO
1043 workloads need only copy that segment, instead of entering
1044 the kernel with a gettimeofday() call. The CPU set aside
1045 for doing these time calls will be excluded from other
1046 uses. Fio will manually clear it from the CPU mask of other
1049 continue_on_error=bool Normally fio will exit the job on the first observed
1050 failure. If this option is set, fio will continue the job when
1051 there is a 'non-fatal error' (EIO or EILSEQ) until the runtime
1052 is exceeded or the I/O size specified is completed. If this
1053 option is used, there are two more stats that are appended,
1054 the total error count and the first error. The error field
1055 given in the stats is the first error that was hit during the
1058 cgroup=str Add job to this control group. If it doesn't exist, it will
1059 be created. The system must have a mounted cgroup blkio
1060 mount point for this to work. If your system doesn't have it
1061 mounted, you can do so with:
1063 # mount -t cgroup -o blkio none /cgroup
1065 cgroup_weight=int Set the weight of the cgroup to this value. See
1066 the documentation that comes with the kernel, allowed values
1067 are in the range of 100..1000.
1069 cgroup_nodelete=bool Normally fio will delete the cgroups it has created after
1070 the job completion. To override this behavior and to leave
1071 cgroups around after the job completion, set cgroup_nodelete=1.
1072 This can be useful if one wants to inspect various cgroup
1073 files after job completion. Default: false
1075 uid=int Instead of running as the invoking user, set the user ID to
1076 this value before the thread/process does any work.
1078 gid=int Set group ID, see uid.
1080 6.0 Interpreting the output
1081 ---------------------------
1083 fio spits out a lot of output. While running, fio will display the
1084 status of the jobs created. An example of that would be:
1086 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1088 The characters inside the square brackets denote the current status of
1089 each thread. The possible values (in typical life cycle order) are:
1093 P Thread setup, but not started.
1095 I Thread initialized, waiting.
1096 p Thread running pre-reading file(s).
1097 R Running, doing sequential reads.
1098 r Running, doing random reads.
1099 W Running, doing sequential writes.
1100 w Running, doing random writes.
1101 M Running, doing mixed sequential reads/writes.
1102 m Running, doing mixed random reads/writes.
1103 F Running, currently waiting for fsync()
1104 V Running, doing verification of written data.
1105 E Thread exited, not reaped by main thread yet.
1108 The other values are fairly self explanatory - number of threads
1109 currently running and doing io, rate of io since last check (read speed
1110 listed first, then write speed), and the estimated completion percentage
1111 and time for the running group. It's impossible to estimate runtime of
1112 the following groups (if any).
1114 When fio is done (or interrupted by ctrl-c), it will show the data for
1115 each thread, group of threads, and disks in that order. For each data
1116 direction, the output looks like:
1118 Client1 (g=0): err= 0:
1119 write: io= 32MB, bw= 666KB/s, runt= 50320msec
1120 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
1121 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
1122 bw (KB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
1123 cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17
1124 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
1125 submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1126 complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
1127 issued r/w: total=0/32768, short=0/0
1128 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
1129 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
1131 The client number is printed, along with the group id and error of that
1132 thread. Below is the io statistics, here for writes. In the order listed,
1135 io= Number of megabytes io performed
1136 bw= Average bandwidth rate
1137 runt= The runtime of that thread
1138 slat= Submission latency (avg being the average, stdev being the
1139 standard deviation). This is the time it took to submit
1140 the io. For sync io, the slat is really the completion
1141 latency, since queue/complete is one operation there. This
1142 value can be in milliseconds or microseconds, fio will choose
1143 the most appropriate base and print that. In the example
1144 above, milliseconds is the best scale.
1145 clat= Completion latency. Same names as slat, this denotes the
1146 time from submission to completion of the io pieces. For
1147 sync io, clat will usually be equal (or very close) to 0,
1148 as the time from submit to complete is basically just
1149 CPU time (io has already been done, see slat explanation).
1150 bw= Bandwidth. Same names as the xlat stats, but also includes
1151 an approximate percentage of total aggregate bandwidth
1152 this thread received in this group. This last value is
1153 only really useful if the threads in this group are on the
1154 same disk, since they are then competing for disk access.
1155 cpu= CPU usage. User and system time, along with the number
1156 of context switches this thread went through, usage of
1157 system and user time, and finally the number of major
1158 and minor page faults.
1159 IO depths= The distribution of io depths over the job life time. The
1160 numbers are divided into powers of 2, so for example the
1161 16= entries includes depths up to that value but higher
1162 than the previous entry. In other words, it covers the
1163 range from 16 to 31.
1164 IO submit= How many pieces of IO were submitting in a single submit
1165 call. Each entry denotes that amount and below, until
1166 the previous entry - eg, 8=100% mean that we submitted
1167 anywhere in between 5-8 ios per submit call.
1168 IO complete= Like the above submit number, but for completions instead.
1169 IO issued= The number of read/write requests issued, and how many
1171 IO latencies= The distribution of IO completion latencies. This is the
1172 time from when IO leaves fio and when it gets completed.
1173 The numbers follow the same pattern as the IO depths,
1174 meaning that 2=1.6% means that 1.6% of the IO completed
1175 within 2 msecs, 20=12.8% means that 12.8% of the IO
1176 took more than 10 msecs, but less than (or equal to) 20 msecs.
1178 After each client has been listed, the group statistics are printed. They
1179 will look like this:
1181 Run status group 0 (all jobs):
1182 READ: io=64MB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
1183 WRITE: io=64MB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
1185 For each data direction, it prints:
1187 io= Number of megabytes io performed.
1188 aggrb= Aggregate bandwidth of threads in this group.
1189 minb= The minimum average bandwidth a thread saw.
1190 maxb= The maximum average bandwidth a thread saw.
1191 mint= The smallest runtime of the threads in that group.
1192 maxt= The longest runtime of the threads in that group.
1194 And finally, the disk statistics are printed. They will look like this:
1196 Disk stats (read/write):
1197 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
1199 Each value is printed for both reads and writes, with reads first. The
1202 ios= Number of ios performed by all groups.
1203 merge= Number of merges io the io scheduler.
1204 ticks= Number of ticks we kept the disk busy.
1205 io_queue= Total time spent in the disk queue.
1206 util= The disk utilization. A value of 100% means we kept the disk
1207 busy constantly, 50% would be a disk idling half of the time.
1213 For scripted usage where you typically want to generate tables or graphs
1214 of the results, fio can output the results in a semicolon separated format.
1215 The format is one long line of values, such as:
1217 2; 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%
1218 ;0.0%;0.0%;0.0%;0.0%;0.0%
1220 To enable terse output, use the --minimal command line option. The first
1221 value is the version of the terse output format. If the output has to
1222 be changed for some reason, this number will be incremented by 1 to
1223 signify that change.
1225 Split up, the format is as follows:
1227 version, jobname, groupid, error
1229 KB IO, bandwidth (KB/sec), runtime (msec)
1230 Submission latency: min, max, mean, deviation
1231 Completion latency: min, max, mean, deviation
1232 Total latency: min, max, mean, deviation
1233 Bw: min, max, aggregate percentage of total, mean, deviation
1235 KB IO, bandwidth (KB/sec), runtime (msec)
1236 Submission latency: min, max, mean, deviation
1237 Completion latency: min, max, mean, deviation
1238 Total latency: min, max, mean, deviation
1239 Bw: min, max, aggregate percentage of total, mean, deviation
1240 CPU usage: user, system, context switches, major faults, minor faults
1241 IO depths: <=1, 2, 4, 8, 16, 32, >=64
1242 IO latencies: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, >=2000
projects / fio.git / blob