1 .TH fio 1 "September 2007" "User Manual"
3 fio \- flexible I/O tester
6 [\fIoptions\fR] [\fIjobfile\fR]...
9 is a tool that will spawn a number of threads or processes doing a
10 particular type of I/O action as specified by the user.
11 The typical use of fio is to write a job file matching the I/O load
12 one wants to simulate.
15 .BI \-\-debug \fR=\fPtype
16 Enable verbose tracing of various fio actions. May be `all' for all types
17 or individual types separated by a comma (eg \-\-debug=io,file). `help' will
18 list all available tracing options.
20 .BI \-\-output \fR=\fPfilename
21 Write output to \fIfilename\fR.
23 .BI \-\-timeout \fR=\fPtimeout
24 Limit run time to \fItimeout\fR seconds.
27 Generate per-job latency logs.
30 Generate per-job bandwidth logs.
33 Print statistics in a terse, semicolon-delimited format.
36 Display version information and exit.
38 .BI \-\-terse\-version \fR=\fPversion
39 Set terse version output format (Current version 3, or older version 2).
42 Display usage information and exit.
44 .BI \-\-cmdhelp \fR=\fPcommand
45 Print help information for \fIcommand\fR. May be `all' for all commands.
47 .BI \-\-enghelp \fR=\fPioengine[,command]
48 List all commands defined by \fIioengine\fR, or print help for \fIcommand\fR defined by \fIioengine\fR.
50 .BI \-\-showcmd \fR=\fPjobfile
51 Convert \fIjobfile\fR to a set of command-line options.
54 Enable read-only safety checks.
56 .BI \-\-eta \fR=\fPwhen
57 Specifies when real-time ETA estimate should be printed. \fIwhen\fR may
58 be one of `always', `never' or `auto'.
61 Turn on safety read-only checks, preventing any attempted write.
63 .BI \-\-section \fR=\fPsec
64 Only run section \fIsec\fR from job file. Multiple of these options can be given, adding more sections to run.
66 .BI \-\-alloc\-size \fR=\fPkb
67 Set the internal smalloc pool size to \fIkb\fP kilobytes.
69 .BI \-\-warnings\-fatal
70 All fio parser warnings are fatal, causing fio to exit with an error.
72 .BI \-\-max\-jobs \fR=\fPnr
73 Set the maximum allowed number of jobs (threads/processes) to suport.
75 .BI \-\-server \fR=\fPargs
76 Start a backend server, with \fIargs\fP specifying what to listen to. See client/server section.
78 .BI \-\-daemonize \fR=\fPpidfile
79 Background a fio server, writing the pid to the given pid file.
81 .BI \-\-client \fR=\fPhost
82 Instead of running the jobs locally, send and run them on the given host.
84 Job files are in `ini' format. They consist of one or more
85 job definitions, which begin with a job name in square brackets and
86 extend to the next job name. The job name can be any ASCII string
87 except `global', which has a special meaning. Following the job name is
88 a sequence of zero or more parameters, one per line, that define the
89 behavior of the job. Any line starting with a `;' or `#' character is
90 considered a comment and ignored.
92 If \fIjobfile\fR is specified as `-', the job file will be read from
95 The global section contains default parameters for jobs specified in the
96 job file. A job is only affected by global sections residing above it,
97 and there may be any number of global sections. Specific job definitions
98 may override any parameter set in global sections.
101 Some parameters may take arguments of a specific type. The types used are:
104 String: a sequence of alphanumeric characters.
107 SI integer: a whole number, possibly containing a suffix denoting the base unit
108 of the value. Accepted suffixes are `k', 'M', 'G', 'T', and 'P', denoting
109 kilo (1024), mega (1024^2), giga (1024^3), tera (1024^4), and peta (1024^5)
110 respectively. The suffix is not case sensitive. If prefixed with '0x', the
111 value is assumed to be base 16 (hexadecimal). A suffix may include a trailing 'b',
112 for instance 'kb' is identical to 'k'. You can specify a base 10 value
113 by using 'KiB', 'MiB', 'GiB', etc. This is useful for disk drives where
114 values are often given in base 10 values. Specifying '30GiB' will get you
118 Boolean: a true or false value. `0' denotes false, `1' denotes true.
121 Integer range: a range of integers specified in the format
122 \fIlower\fR:\fIupper\fR or \fIlower\fR\-\fIupper\fR. \fIlower\fR and
123 \fIupper\fR may contain a suffix as described above. If an option allows two
124 sets of ranges, they are separated with a `,' or `/' character. For example:
128 List of floating numbers: A list of floating numbers, separated by
133 May be used to override the job name. On the command line, this parameter
134 has the special purpose of signalling the start of a new job.
136 .BI description \fR=\fPstr
137 Human-readable description of the job. It is printed when the job is run, but
138 otherwise has no special purpose.
140 .BI directory \fR=\fPstr
141 Prefix filenames with this directory. Used to place files in a location other
144 .BI filename \fR=\fPstr
146 normally makes up a file name based on the job name, thread number, and file
147 number. If you want to share files between threads in a job or several jobs,
148 specify a \fIfilename\fR for each of them to override the default.
149 If the I/O engine is file-based, you can specify
150 a number of files by separating the names with a `:' character. `\-' is a
151 reserved name, meaning stdin or stdout, depending on the read/write direction
154 .BI lockfile \fR=\fPstr
155 Fio defaults to not locking any files before it does IO to them. If a file or
156 file descriptor is shared, fio can serialize IO to that file to make the end
157 result consistent. This is usual for emulating real workloads that share files.
163 No locking. This is the default.
166 Only one thread or process may do IO at the time, excluding all others.
169 Read-write locking on the file. Many readers may access the file at the same
170 time, but writes get exclusive access.
173 The option may be post-fixed with a lock batch number. If set, then each
174 thread/process may do that amount of IOs to the file before giving up the lock.
175 Since lock acquisition is expensive, batching the lock/unlocks will speed up IO.
178 .BI opendir \fR=\fPstr
179 Recursively open any files below directory \fIstr\fR.
181 .BI readwrite \fR=\fPstr "\fR,\fP rw" \fR=\fPstr
182 Type of I/O pattern. Accepted values are:
199 Mixed sequential reads and writes.
202 Mixed random reads and writes.
205 For mixed I/O, the default split is 50/50. For certain types of io the result
206 may still be skewed a bit, since the speed may be different. It is possible to
207 specify a number of IO's to do before getting a new offset, this is one by
208 appending a `:\fI<nr>\fR to the end of the string given. For a random read, it
209 would look like \fBrw=randread:8\fR for passing in an offset modifier with a
210 value of 8. If the postfix is used with a sequential IO pattern, then the value
211 specified will be added to the generated offset for each IO. For instance,
212 using \fBrw=write:4k\fR will skip 4k for every write. It turns sequential IO
213 into sequential IO with holes. See the \fBrw_sequencer\fR option.
216 .BI rw_sequencer \fR=\fPstr
217 If an offset modifier is given by appending a number to the \fBrw=<str>\fR line,
218 then this option controls how that number modifies the IO offset being
219 generated. Accepted values are:
224 Generate sequential offset
227 Generate the same offset
230 \fBsequential\fR is only useful for random IO, where fio would normally
231 generate a new random offset for every IO. If you append eg 8 to randread, you
232 would get a new random offset for every 8 IO's. The result would be a seek for
233 only every 8 IO's, instead of for every IO. Use \fBrw=randread:8\fR to specify
234 that. As sequential IO is already sequential, setting \fBsequential\fR for that
235 would not result in any differences. \fBidentical\fR behaves in a similar
236 fashion, except it sends the same offset 8 number of times before generating a
241 .BI kb_base \fR=\fPint
242 The base unit for a kilobyte. The defacto base is 2^10, 1024. Storage
243 manufacturers like to use 10^3 or 1000 as a base ten unit instead, for obvious
244 reasons. Allow values are 1024 or 1000, with 1024 being the default.
246 .BI randrepeat \fR=\fPbool
247 Seed the random number generator in a predictable way so results are repeatable
248 across runs. Default: true.
250 .BI use_os_rand \fR=\fPbool
251 Fio can either use the random generator supplied by the OS to generator random
252 offsets, or it can use it's own internal generator (based on Tausworthe).
253 Default is to use the internal generator, which is often of better quality and
254 faster. Default: false.
256 .BI fallocate \fR=\fPstr
257 Whether pre-allocation is performed when laying down files. Accepted values
263 Do not pre-allocate space.
266 Pre-allocate via posix_fallocate().
269 Pre-allocate via fallocate() with FALLOC_FL_KEEP_SIZE set.
272 Backward-compatible alias for 'none'.
275 Backward-compatible alias for 'posix'.
278 May not be available on all supported platforms. 'keep' is only
279 available on Linux. If using ZFS on Solaris this must be set to 'none'
280 because ZFS doesn't support it. Default: 'posix'.
283 .BI fadvise_hint \fR=\fPbool
284 Use of \fIposix_fadvise\fR\|(2) to advise the kernel what I/O patterns
285 are likely to be issued. Default: true.
288 Total size of I/O for this job. \fBfio\fR will run until this many bytes have
289 been transfered, unless limited by other options (\fBruntime\fR, for instance).
290 Unless \fBnrfiles\fR and \fBfilesize\fR options are given, this amount will be
291 divided between the available files for the job. If not set, fio will use the
292 full size of the given files or devices. If the the files do not exist, size
293 must be given. It is also possible to give size as a percentage between 1 and
294 100. If size=20% is given, fio will use 20% of the full size of the given files
297 .BI fill_device \fR=\fPbool "\fR,\fB fill_fs" \fR=\fPbool
298 Sets size to something really large and waits for ENOSPC (no space left on
299 device) as the terminating condition. Only makes sense with sequential write.
300 For a read workload, the mount point will be filled first then IO started on
301 the result. This option doesn't make sense if operating on a raw device node,
302 since the size of that is already known by the file system. Additionally,
303 writing beyond end-of-device will not return ENOSPC there.
305 .BI filesize \fR=\fPirange
306 Individual file sizes. May be a range, in which case \fBfio\fR will select sizes
307 for files at random within the given range, limited to \fBsize\fR in total (if
308 that is given). If \fBfilesize\fR is not specified, each created file is the
311 .BI blocksize \fR=\fPint[,int] "\fR,\fB bs" \fR=\fPint[,int]
312 Block size for I/O units. Default: 4k. Values for reads and writes can be
313 specified separately in the format \fIread\fR,\fIwrite\fR, either of
314 which may be empty to leave that value at its default.
316 .BI blocksize_range \fR=\fPirange[,irange] "\fR,\fB bsrange" \fR=\fPirange[,irange]
317 Specify a range of I/O block sizes. The issued I/O unit will always be a
318 multiple of the minimum size, unless \fBblocksize_unaligned\fR is set. Applies
319 to both reads and writes if only one range is given, but can be specified
320 separately with a comma seperating the values. Example: bsrange=1k-4k,2k-8k.
321 Also (see \fBblocksize\fR).
323 .BI bssplit \fR=\fPstr
324 This option allows even finer grained control of the block sizes issued,
325 not just even splits between them. With this option, you can weight various
326 block sizes for exact control of the issued IO for a job that has mixed
327 block sizes. The format of the option is bssplit=blocksize/percentage,
328 optionally adding as many definitions as needed separated by a colon.
329 Example: bssplit=4k/10:64k/50:32k/40 would issue 50% 64k blocks, 10% 4k
330 blocks and 40% 32k blocks. \fBbssplit\fR also supports giving separate
331 splits to reads and writes. The format is identical to what the
332 \fBbs\fR option accepts, the read and write parts are separated with a
335 .B blocksize_unaligned\fR,\fP bs_unaligned
336 If set, any size in \fBblocksize_range\fR may be used. This typically won't
337 work with direct I/O, as that normally requires sector alignment.
339 .BI blockalign \fR=\fPint[,int] "\fR,\fB ba" \fR=\fPint[,int]
340 At what boundary to align random IO offsets. Defaults to the same as 'blocksize'
341 the minimum blocksize given. Minimum alignment is typically 512b
342 for using direct IO, though it usually depends on the hardware block size.
343 This option is mutually exclusive with using a random map for files, so it
344 will turn off that option.
347 Initialise buffers with all zeros. Default: fill buffers with random data.
350 If this option is given, fio will refill the IO buffers on every submit. The
351 default is to only fill it at init time and reuse that data. Only makes sense
352 if zero_buffers isn't specified, naturally. If data verification is enabled,
353 refill_buffers is also automatically enabled.
355 .BI scramble_buffers \fR=\fPbool
356 If \fBrefill_buffers\fR is too costly and the target is using data
357 deduplication, then setting this option will slightly modify the IO buffer
358 contents to defeat normal de-dupe attempts. This is not enough to defeat
359 more clever block compression attempts, but it will stop naive dedupe
360 of blocks. Default: true.
362 .BI buffer_compress_percentage \fR=\fPint
363 If this is set, then fio will attempt to provide IO buffer content (on WRITEs)
364 that compress to the specified level. Fio does this by providing a mix of
365 random data and zeroes. Note that this is per block size unit, for file/disk
366 wide compression level that matches this setting, you'll also want to set
367 \fBrefill_buffers\fR.
369 .BI buffer_compress_chunk \fR=\fPint
370 See \fBbuffer_compress_percentage\fR. This setting allows fio to manage how
371 big the ranges of random data and zeroed data is. Without this set, fio will
372 provide \fBbuffer_compress_percentage\fR of blocksize random data, followed by
373 the remaining zeroed. With this set to some chunk size smaller than the block
374 size, fio can alternate random and zeroed data throughout the IO buffer.
376 .BI nrfiles \fR=\fPint
377 Number of files to use for this job. Default: 1.
379 .BI openfiles \fR=\fPint
380 Number of files to keep open at the same time. Default: \fBnrfiles\fR.
382 .BI file_service_type \fR=\fPstr
383 Defines how files to service are selected. The following types are defined:
388 Choose a file at random
391 Round robin over open files (default).
393 Do each file in the set sequentially.
396 The number of I/Os to issue before switching a new file can be specified by
397 appending `:\fIint\fR' to the service type.
400 .BI ioengine \fR=\fPstr
401 Defines how the job issues I/O. The following types are defined:
406 Basic \fIread\fR\|(2) or \fIwrite\fR\|(2) I/O. \fIfseek\fR\|(2) is used to
407 position the I/O location.
410 Basic \fIpread\fR\|(2) or \fIpwrite\fR\|(2) I/O.
413 Basic \fIreadv\fR\|(2) or \fIwritev\fR\|(2) I/O. Will emulate queuing by
414 coalescing adjacents IOs into a single submission.
417 Linux native asynchronous I/O. This ioengine defines engine specific options.
420 POSIX asynchronous I/O using \fIaio_read\fR\|(3) and \fIaio_write\fR\|(3).
423 Solaris native asynchronous I/O.
426 Windows native asynchronous I/O.
429 File is memory mapped with \fImmap\fR\|(2) and data copied using
433 \fIsplice\fR\|(2) is used to transfer the data and \fIvmsplice\fR\|(2) to
434 transfer data from user-space to the kernel.
437 Use the syslet system calls to make regular read/write asynchronous.
440 SCSI generic sg v3 I/O. May be either synchronous using the SG_IO ioctl, or if
441 the target is an sg character device, we use \fIread\fR\|(2) and
442 \fIwrite\fR\|(2) for asynchronous I/O.
445 Doesn't transfer any data, just pretends to. Mainly used to exercise \fBfio\fR
446 itself and for debugging and testing purposes.
449 Transfer over the network. The protocol to be used can be defined with the
450 \fBprotocol\fR parameter. Depending on the protocol, \fBfilename\fR,
451 \fBhostname\fR, \fBport\fR, or \fBlisten\fR must be specified.
452 This ioengine defines engine specific options.
455 Like \fBnet\fR, but uses \fIsplice\fR\|(2) and \fIvmsplice\fR\|(2) to map data
456 and send/receive. This ioengine defines engine specific options.
459 Doesn't transfer any data, but burns CPU cycles according to \fBcpuload\fR and
460 \fBcpucycles\fR parameters.
463 The GUASI I/O engine is the Generic Userspace Asynchronous Syscall Interface
464 approach to asycnronous I/O.
466 See <http://www.xmailserver.org/guasi\-lib.html>.
469 The RDMA I/O engine supports both RDMA memory semantics (RDMA_WRITE/RDMA_READ)
470 and channel semantics (Send/Recv) for the InfiniBand, RoCE and iWARP protocols.
473 Loads an external I/O engine object file. Append the engine filename as
478 .BI iodepth \fR=\fPint
479 Number of I/O units to keep in flight against the file. Note that increasing
480 iodepth beyond 1 will not affect synchronous ioengines (except for small
481 degress when verify_async is in use). Even async engines my impose OS
482 restrictions causing the desired depth not to be achieved. This may happen on
483 Linux when using libaio and not setting \fBdirect\fR=1, since buffered IO is
484 not async on that OS. Keep an eye on the IO depth distribution in the
485 fio output to verify that the achieved depth is as expected. Default: 1.
487 .BI iodepth_batch \fR=\fPint
488 Number of I/Os to submit at once. Default: \fBiodepth\fR.
490 .BI iodepth_batch_complete \fR=\fPint
491 This defines how many pieces of IO to retrieve at once. It defaults to 1 which
492 means that we'll ask for a minimum of 1 IO in the retrieval process from the
493 kernel. The IO retrieval will go on until we hit the limit set by
494 \fBiodepth_low\fR. If this variable is set to 0, then fio will always check for
495 completed events before queuing more IO. This helps reduce IO latency, at the
496 cost of more retrieval system calls.
498 .BI iodepth_low \fR=\fPint
499 Low watermark indicating when to start filling the queue again. Default:
502 .BI direct \fR=\fPbool
503 If true, use non-buffered I/O (usually O_DIRECT). Default: false.
505 .BI buffered \fR=\fPbool
506 If true, use buffered I/O. This is the opposite of the \fBdirect\fR parameter.
509 .BI offset \fR=\fPint
510 Offset in the file to start I/O. Data before the offset will not be touched.
512 .BI offset_increment \fR=\fPint
513 If this is provided, then the real offset becomes the
514 offset + offset_increment * thread_number, where the thread number is a counter
515 that starts at 0 and is incremented for each job. This option is useful if
516 there are several jobs which are intended to operate on a file in parallel in
517 disjoint segments, with even spacing between the starting points.
520 How many I/Os to perform before issuing an \fBfsync\fR\|(2) of dirty data. If
521 0, don't sync. Default: 0.
523 .BI fdatasync \fR=\fPint
524 Like \fBfsync\fR, but uses \fBfdatasync\fR\|(2) instead to only sync the
525 data parts of the file. Default: 0.
527 .BI sync_file_range \fR=\fPstr:int
528 Use sync_file_range() for every \fRval\fP number of write operations. Fio will
529 track range of writes that have happened since the last sync_file_range() call.
530 \fRstr\fP can currently be one or more of:
534 SYNC_FILE_RANGE_WAIT_BEFORE
537 SYNC_FILE_RANGE_WRITE
540 SYNC_FILE_RANGE_WRITE
544 So if you do sync_file_range=wait_before,write:8, fio would use
545 \fBSYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE\fP for every 8 writes.
546 Also see the sync_file_range(2) man page. This option is Linux specific.
548 .BI overwrite \fR=\fPbool
549 If writing, setup the file first and do overwrites. Default: false.
551 .BI end_fsync \fR=\fPbool
552 Sync file contents when job exits. Default: false.
554 .BI fsync_on_close \fR=\fPbool
555 If true, sync file contents on close. This differs from \fBend_fsync\fR in that
556 it will happen on every close, not just at the end of the job. Default: false.
558 .BI rwmixread \fR=\fPint
559 Percentage of a mixed workload that should be reads. Default: 50.
561 .BI rwmixwrite \fR=\fPint
562 Percentage of a mixed workload that should be writes. If \fBrwmixread\fR and
563 \fBrwmixwrite\fR are given and do not sum to 100%, the latter of the two
564 overrides the first. This may interfere with a given rate setting, if fio is
565 asked to limit reads or writes to a certain rate. If that is the case, then
566 the distribution may be skewed. Default: 50.
569 Normally \fBfio\fR will cover every block of the file when doing random I/O. If
570 this parameter is given, a new offset will be chosen without looking at past
571 I/O history. This parameter is mutually exclusive with \fBverify\fR.
573 .BI softrandommap \fR=\fPbool
574 See \fBnorandommap\fR. If fio runs with the random block map enabled and it
575 fails to allocate the map, if this option is set it will continue without a
576 random block map. As coverage will not be as complete as with random maps, this
577 option is disabled by default.
580 Run job with given nice value. See \fInice\fR\|(2).
583 Set I/O priority value of this job between 0 (highest) and 7 (lowest). See
586 .BI prioclass \fR=\fPint
587 Set I/O priority class. See \fIionice\fR\|(1).
589 .BI thinktime \fR=\fPint
590 Stall job for given number of microseconds between issuing I/Os.
592 .BI thinktime_spin \fR=\fPint
593 Pretend to spend CPU time for given number of microseconds, sleeping the rest
594 of the time specified by \fBthinktime\fR. Only valid if \fBthinktime\fR is set.
596 .BI thinktime_blocks \fR=\fPint
597 Number of blocks to issue before waiting \fBthinktime\fR microseconds.
601 Cap bandwidth used by this job. The number is in bytes/sec, the normal postfix
602 rules apply. You can use \fBrate\fR=500k to limit reads and writes to 500k each,
603 or you can specify read and writes separately. Using \fBrate\fR=1m,500k would
604 limit reads to 1MB/sec and writes to 500KB/sec. Capping only reads or writes
605 can be done with \fBrate\fR=,500k or \fBrate\fR=500k,. The former will only
606 limit writes (to 500KB/sec), the latter will only limit reads.
608 .BI ratemin \fR=\fPint
609 Tell \fBfio\fR to do whatever it can to maintain at least the given bandwidth.
610 Failing to meet this requirement will cause the job to exit. The same format
611 as \fBrate\fR is used for read vs write separation.
613 .BI rate_iops \fR=\fPint
614 Cap the bandwidth to this number of IOPS. Basically the same as rate, just
615 specified independently of bandwidth. The same format as \fBrate\fR is used for
616 read vs write seperation. If \fBblocksize\fR is a range, the smallest block
617 size is used as the metric.
619 .BI rate_iops_min \fR=\fPint
620 If this rate of I/O is not met, the job will exit. The same format as \fBrate\fR
621 is used for read vs write seperation.
623 .BI ratecycle \fR=\fPint
624 Average bandwidth for \fBrate\fR and \fBratemin\fR over this number of
625 milliseconds. Default: 1000ms.
627 .BI cpumask \fR=\fPint
628 Set CPU affinity for this job. \fIint\fR is a bitmask of allowed CPUs the job
629 may run on. See \fBsched_setaffinity\fR\|(2).
631 .BI cpus_allowed \fR=\fPstr
632 Same as \fBcpumask\fR, but allows a comma-delimited list of CPU numbers.
634 .BI startdelay \fR=\fPint
635 Delay start of job for the specified number of seconds.
637 .BI runtime \fR=\fPint
638 Terminate processing after the specified number of seconds.
641 If given, run for the specified \fBruntime\fR duration even if the files are
642 completely read or written. The same workload will be repeated as many times
643 as \fBruntime\fR allows.
645 .BI ramp_time \fR=\fPint
646 If set, fio will run the specified workload for this amount of time before
647 logging any performance numbers. Useful for letting performance settle before
648 logging results, thus minimizing the runtime required for stable results. Note
649 that the \fBramp_time\fR is considered lead in time for a job, thus it will
650 increase the total runtime if a special timeout or runtime is specified.
652 .BI invalidate \fR=\fPbool
653 Invalidate buffer-cache for the file prior to starting I/O. Default: true.
656 Use synchronous I/O for buffered writes. For the majority of I/O engines,
657 this means using O_SYNC. Default: false.
659 .BI iomem \fR=\fPstr "\fR,\fP mem" \fR=\fPstr
660 Allocation method for I/O unit buffer. Allowed values are:
665 Allocate memory with \fImalloc\fR\|(3).
668 Use shared memory buffers allocated through \fIshmget\fR\|(2).
671 Same as \fBshm\fR, but use huge pages as backing.
674 Use \fImmap\fR\|(2) for allocation. Uses anonymous memory unless a filename
675 is given after the option in the format `:\fIfile\fR'.
678 Same as \fBmmap\fR, but use huge files as backing.
681 The amount of memory allocated is the maximum allowed \fBblocksize\fR for the
682 job multiplied by \fBiodepth\fR. For \fBshmhuge\fR or \fBmmaphuge\fR to work,
683 the system must have free huge pages allocated. \fBmmaphuge\fR also needs to
684 have hugetlbfs mounted, and \fIfile\fR must point there. At least on Linux,
685 huge pages must be manually allocated. See \fB/proc/sys/vm/nr_hugehages\fR
686 and the documentation for that. Normally you just need to echo an appropriate
687 number, eg echoing 8 will ensure that the OS has 8 huge pages ready for
691 .BI iomem_align \fR=\fPint "\fR,\fP mem_align" \fR=\fPint
692 This indiciates the memory alignment of the IO memory buffers. Note that the
693 given alignment is applied to the first IO unit buffer, if using \fBiodepth\fR
694 the alignment of the following buffers are given by the \fBbs\fR used. In
695 other words, if using a \fBbs\fR that is a multiple of the page sized in the
696 system, all buffers will be aligned to this value. If using a \fBbs\fR that
697 is not page aligned, the alignment of subsequent IO memory buffers is the
698 sum of the \fBiomem_align\fR and \fBbs\fR used.
700 .BI hugepage\-size \fR=\fPint
701 Defines the size of a huge page. Must be at least equal to the system setting.
702 Should be a multiple of 1MB. Default: 4MB.
705 Terminate all jobs when one finishes. Default: wait for each job to finish.
707 .BI bwavgtime \fR=\fPint
708 Average bandwidth calculations over the given time in milliseconds. Default:
711 .BI iopsavgtime \fR=\fPint
712 Average IOPS calculations over the given time in milliseconds. Default:
715 .BI create_serialize \fR=\fPbool
716 If true, serialize file creation for the jobs. Default: true.
718 .BI create_fsync \fR=\fPbool
719 \fIfsync\fR\|(2) data file after creation. Default: true.
721 .BI create_on_open \fR=\fPbool
722 If true, the files are not created until they are opened for IO by the job.
724 .BI create_only \fR=\fPbool
725 If true, fio will only run the setup phase of the job. If files need to be
726 laid out or updated on disk, only that will be done. The actual job contents
729 .BI pre_read \fR=\fPbool
730 If this is given, files will be pre-read into memory before starting the given
731 IO operation. This will also clear the \fR \fBinvalidate\fR flag, since it is
732 pointless to pre-read and then drop the cache. This will only work for IO
733 engines that are seekable, since they allow you to read the same data
734 multiple times. Thus it will not work on eg network or splice IO.
736 .BI unlink \fR=\fPbool
737 Unlink job files when done. Default: false.
740 Specifies the number of iterations (runs of the same workload) of this job.
743 .BI do_verify \fR=\fPbool
744 Run the verify phase after a write phase. Only valid if \fBverify\fR is set.
747 .BI verify \fR=\fPstr
748 Method of verifying file contents after each iteration of the job. Allowed
753 .B md5 crc16 crc32 crc32c crc32c-intel crc64 crc7 sha256 sha512 sha1
754 Store appropriate checksum in the header of each block. crc32c-intel is
755 hardware accelerated SSE4.2 driven, falls back to regular crc32c if
756 not supported by the system.
759 Write extra information about each I/O (timestamp, block number, etc.). The
760 block number is verified. See \fBverify_pattern\fR as well.
763 Pretend to verify. Used for testing internals.
766 This option can be used for repeated burn-in tests of a system to make sure
767 that the written data is also correctly read back. If the data direction given
768 is a read or random read, fio will assume that it should verify a previously
769 written file. If the data direction includes any form of write, the verify will
770 be of the newly written data.
773 .BI verify_sort \fR=\fPbool
774 If true, written verify blocks are sorted if \fBfio\fR deems it to be faster to
775 read them back in a sorted manner. Default: true.
777 .BI verify_offset \fR=\fPint
778 Swap the verification header with data somewhere else in the block before
779 writing. It is swapped back before verifying.
781 .BI verify_interval \fR=\fPint
782 Write the verification header for this number of bytes, which should divide
783 \fBblocksize\fR. Default: \fBblocksize\fR.
785 .BI verify_pattern \fR=\fPstr
786 If set, fio will fill the io buffers with this pattern. Fio defaults to filling
787 with totally random bytes, but sometimes it's interesting to fill with a known
788 pattern for io verification purposes. Depending on the width of the pattern,
789 fio will fill 1/2/3/4 bytes of the buffer at the time(it can be either a
790 decimal or a hex number). The verify_pattern if larger than a 32-bit quantity
791 has to be a hex number that starts with either "0x" or "0X". Use with
794 .BI verify_fatal \fR=\fPbool
795 If true, exit the job on the first observed verification failure. Default:
798 .BI verify_dump \fR=\fPbool
799 If set, dump the contents of both the original data block and the data block we
800 read off disk to files. This allows later analysis to inspect just what kind of
801 data corruption occurred. Off by default.
803 .BI verify_async \fR=\fPint
804 Fio will normally verify IO inline from the submitting thread. This option
805 takes an integer describing how many async offload threads to create for IO
806 verification instead, causing fio to offload the duty of verifying IO contents
807 to one or more separate threads. If using this offload option, even sync IO
808 engines can benefit from using an \fBiodepth\fR setting higher than 1, as it
809 allows them to have IO in flight while verifies are running.
811 .BI verify_async_cpus \fR=\fPstr
812 Tell fio to set the given CPU affinity on the async IO verification threads.
813 See \fBcpus_allowed\fP for the format used.
815 .BI verify_backlog \fR=\fPint
816 Fio will normally verify the written contents of a job that utilizes verify
817 once that job has completed. In other words, everything is written then
818 everything is read back and verified. You may want to verify continually
819 instead for a variety of reasons. Fio stores the meta data associated with an
820 IO block in memory, so for large verify workloads, quite a bit of memory would
821 be used up holding this meta data. If this option is enabled, fio will write
822 only N blocks before verifying these blocks.
824 .BI verify_backlog_batch \fR=\fPint
825 Control how many blocks fio will verify if verify_backlog is set. If not set,
826 will default to the value of \fBverify_backlog\fR (meaning the entire queue is
827 read back and verified). If \fBverify_backlog_batch\fR is less than
828 \fBverify_backlog\fR then not all blocks will be verified, if
829 \fBverify_backlog_batch\fR is larger than \fBverify_backlog\fR, some blocks
830 will be verified more than once.
832 .B stonewall "\fR,\fP wait_for_previous"
833 Wait for preceding jobs in the job file to exit before starting this one.
834 \fBstonewall\fR implies \fBnew_group\fR.
837 Start a new reporting group. If not given, all jobs in a file will be part
838 of the same reporting group, unless separated by a stonewall.
840 .BI numjobs \fR=\fPint
841 Number of clones (processes/threads performing the same workload) of this job.
845 If set, display per-group reports instead of per-job when \fBnumjobs\fR is
849 Use threads created with \fBpthread_create\fR\|(3) instead of processes created
850 with \fBfork\fR\|(2).
852 .BI zonesize \fR=\fPint
853 Divide file into zones of the specified size in bytes. See \fBzoneskip\fR.
855 .BI zoneskip \fR=\fPint
856 Skip the specified number of bytes when \fBzonesize\fR bytes of data have been
859 .BI write_iolog \fR=\fPstr
860 Write the issued I/O patterns to the specified file. Specify a separate file
861 for each job, otherwise the iologs will be interspersed and the file may be
864 .BI read_iolog \fR=\fPstr
865 Replay the I/O patterns contained in the specified file generated by
866 \fBwrite_iolog\fR, or may be a \fBblktrace\fR binary file.
868 .BI replay_no_stall \fR=\fPint
869 While replaying I/O patterns using \fBread_iolog\fR the default behavior
870 attempts to respect timing information between I/Os. Enabling
871 \fBreplay_no_stall\fR causes I/Os to be replayed as fast as possible while
872 still respecting ordering.
874 .BI replay_redirect \fR=\fPstr
875 While replaying I/O patterns using \fBread_iolog\fR the default behavior
876 is to replay the IOPS onto the major/minor device that each IOP was recorded
877 from. Setting \fBreplay_redirect\fR causes all IOPS to be replayed onto the
878 single specified device regardless of the device it was recorded from.
880 .BI write_bw_log \fR=\fPstr
881 If given, write a bandwidth log of the jobs in this job file. Can be used to
882 store data of the bandwidth of the jobs in their lifetime. The included
883 fio_generate_plots script uses gnuplot to turn these text files into nice
884 graphs. See \fBwrite_log_log\fR for behaviour of given filename. For this
885 option, the postfix is _bw.log.
887 .BI write_lat_log \fR=\fPstr
888 Same as \fBwrite_bw_log\fR, but writes I/O completion latencies. If no
889 filename is given with this option, the default filename of "jobname_type.log"
890 is used. Even if the filename is given, fio will still append the type of log.
892 .BI write_iops_log \fR=\fPstr
893 Same as \fBwrite_bw_log\fR, but writes IOPS. If no filename is given with this
894 option, the default filename of "jobname_type.log" is used. Even if the
895 filename is given, fio will still append the type of log.
897 .BI log_avg_msec \fR=\fPint
898 By default, fio will log an entry in the iops, latency, or bw log for every
899 IO that completes. When writing to the disk log, that can quickly grow to a
900 very large size. Setting this option makes fio average the each log entry
901 over the specified period of time, reducing the resolution of the log.
904 .BI disable_lat \fR=\fPbool
905 Disable measurements of total latency numbers. Useful only for cutting
906 back the number of calls to gettimeofday, as that does impact performance at
907 really high IOPS rates. Note that to really get rid of a large amount of these
908 calls, this option must be used with disable_slat and disable_bw as well.
910 .BI disable_clat \fR=\fPbool
911 Disable measurements of completion latency numbers. See \fBdisable_lat\fR.
913 .BI disable_slat \fR=\fPbool
914 Disable measurements of submission latency numbers. See \fBdisable_lat\fR.
916 .BI disable_bw_measurement \fR=\fPbool
917 Disable measurements of throughput/bandwidth numbers. See \fBdisable_lat\fR.
919 .BI lockmem \fR=\fPint
920 Pin the specified amount of memory with \fBmlock\fR\|(2). Can be used to
921 simulate a smaller amount of memory.
923 .BI exec_prerun \fR=\fPstr
924 Before running the job, execute the specified command with \fBsystem\fR\|(3).
926 .BI exec_postrun \fR=\fPstr
927 Same as \fBexec_prerun\fR, but the command is executed after the job completes.
929 .BI ioscheduler \fR=\fPstr
930 Attempt to switch the device hosting the file to the specified I/O scheduler.
932 .BI cpuload \fR=\fPint
933 If the job is a CPU cycle-eater, attempt to use the specified percentage of
936 .BI cpuchunks \fR=\fPint
937 If the job is a CPU cycle-eater, split the load into cycles of the
938 given time in milliseconds.
940 .BI disk_util \fR=\fPbool
941 Generate disk utilization statistics if the platform supports it. Default: true.
943 .BI gtod_reduce \fR=\fPbool
944 Enable all of the gettimeofday() reducing options (disable_clat, disable_slat,
945 disable_bw) plus reduce precision of the timeout somewhat to really shrink the
946 gettimeofday() call count. With this option enabled, we only do about 0.4% of
947 the gtod() calls we would have done if all time keeping was enabled.
949 .BI gtod_cpu \fR=\fPint
950 Sometimes it's cheaper to dedicate a single thread of execution to just getting
951 the current time. Fio (and databases, for instance) are very intensive on
952 gettimeofday() calls. With this option, you can set one CPU aside for doing
953 nothing but logging current time to a shared memory location. Then the other
954 threads/processes that run IO workloads need only copy that segment, instead of
955 entering the kernel with a gettimeofday() call. The CPU set aside for doing
956 these time calls will be excluded from other uses. Fio will manually clear it
957 from the CPU mask of other jobs.
959 .BI cgroup \fR=\fPstr
960 Add job to this control group. If it doesn't exist, it will be created.
961 The system must have a mounted cgroup blkio mount point for this to work. If
962 your system doesn't have it mounted, you can do so with:
964 # mount \-t cgroup \-o blkio none /cgroup
966 .BI cgroup_weight \fR=\fPint
967 Set the weight of the cgroup to this value. See the documentation that comes
968 with the kernel, allowed values are in the range of 100..1000.
970 .BI cgroup_nodelete \fR=\fPbool
971 Normally fio will delete the cgroups it has created after the job completion.
972 To override this behavior and to leave cgroups around after the job completion,
973 set cgroup_nodelete=1. This can be useful if one wants to inspect various
974 cgroup files after job completion. Default: false
977 Instead of running as the invoking user, set the user ID to this value before
978 the thread/process does any work.
981 Set group ID, see \fBuid\fR.
983 .BI flow_id \fR=\fPint
984 The ID of the flow. If not specified, it defaults to being a global flow. See
988 Weight in token-based flow control. If this value is used, then there is a
989 \fBflow counter\fR which is used to regulate the proportion of activity between
990 two or more jobs. fio attempts to keep this flow counter near zero. The
991 \fBflow\fR parameter stands for how much should be added or subtracted to the
992 flow counter on each iteration of the main I/O loop. That is, if one job has
993 \fBflow=8\fR and another job has \fBflow=-1\fR, then there will be a roughly
994 1:8 ratio in how much one runs vs the other.
996 .BI flow_watermark \fR=\fPint
997 The maximum value that the absolute value of the flow counter is allowed to
998 reach before the job must wait for a lower value of the counter.
1000 .BI flow_sleep \fR=\fPint
1001 The period of time, in microseconds, to wait after the flow watermark has been
1002 exceeded before retrying operations
1004 .BI clat_percentiles \fR=\fPbool
1005 Enable the reporting of percentiles of completion latencies.
1007 .BI percentile_list \fR=\fPfloat_list
1008 Overwrite the default list of percentiles for completion
1009 latencies. Each number is a floating number in the range (0,100], and
1010 the maximum length of the list is 20. Use ':' to separate the
1011 numbers. For example, \-\-percentile_list=99.5:99.9 will cause fio to
1012 report the values of completion latency below which 99.5% and 99.9% of
1013 the observed latencies fell, respectively.
1014 .SS "Ioengine Parameters List"
1015 Some parameters are only valid when a specific ioengine is in use. These are
1016 used identically to normal parameters, with the caveat that when used on the
1017 command line, the must come after the ioengine that defines them is selected.
1019 .BI (libaio)userspace_reap
1020 Normally, with the libaio engine in use, fio will use
1021 the io_getevents system call to reap newly returned events.
1022 With this flag turned on, the AIO ring will be read directly
1023 from user-space to reap events. The reaping mode is only
1024 enabled when polling for a minimum of 0 events (eg when
1025 iodepth_batch_complete=0).
1027 .BI (net,netsplice)hostname \fR=\fPstr
1028 The host name or IP address to use for TCP or UDP based IO.
1029 If the job is a TCP listener or UDP reader, the hostname is not
1030 used and must be omitted.
1032 .BI (net,netsplice)port \fR=\fPint
1033 The TCP or UDP port to bind to or connect to.
1035 .BI (net,netsplice)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr
1036 The network protocol to use. Accepted values are:
1041 Transmission control protocol
1044 Unreliable datagram protocol
1050 When the protocol is TCP or UDP, the port must also be given,
1051 as well as the hostname if the job is a TCP listener or UDP
1052 reader. For unix sockets, the normal filename option should be
1053 used and the port is invalid.
1056 .BI (net,netsplice)listen
1057 For TCP network connections, tell fio to listen for incoming
1058 connections rather than initiating an outgoing connection. The
1059 hostname must be omitted if this option is used.
1061 While running, \fBfio\fR will display the status of the created jobs. For
1065 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
1068 The characters in the first set of brackets denote the current status of each
1069 threads. The possible values are:
1075 Setup but not started.
1081 Initialized, waiting.
1084 Running, doing sequential reads.
1087 Running, doing random reads.
1090 Running, doing sequential writes.
1093 Running, doing random writes.
1096 Running, doing mixed sequential reads/writes.
1099 Running, doing mixed random reads/writes.
1102 Running, currently waiting for \fBfsync\fR\|(2).
1105 Running, verifying written data.
1108 Exited, not reaped by main thread.
1111 Exited, thread reaped.
1115 The second set of brackets shows the estimated completion percentage of
1116 the current group. The third set shows the read and write I/O rate,
1117 respectively. Finally, the estimated run time of the job is displayed.
1119 When \fBfio\fR completes (or is interrupted by Ctrl-C), it will show data
1120 for each thread, each group of threads, and each disk, in that order.
1122 Per-thread statistics first show the threads client number, group-id, and
1123 error code. The remaining figures are as follows:
1127 Number of megabytes of I/O performed.
1130 Average data rate (bandwidth).
1136 Submission latency minimum, maximum, average and standard deviation. This is
1137 the time it took to submit the I/O.
1140 Completion latency minimum, maximum, average and standard deviation. This
1141 is the time between submission and completion.
1144 Bandwidth minimum, maximum, percentage of aggregate bandwidth received, average
1145 and standard deviation.
1148 CPU usage statistics. Includes user and system time, number of context switches
1149 this thread went through and number of major and minor page faults.
1152 Distribution of I/O depths. Each depth includes everything less than (or equal)
1153 to it, but greater than the previous depth.
1156 Number of read/write requests issued, and number of short read/write requests.
1159 Distribution of I/O completion latencies. The numbers follow the same pattern
1163 The group statistics show:
1168 Number of megabytes I/O performed.
1171 Aggregate bandwidth of threads in the group.
1174 Minimum average bandwidth a thread saw.
1177 Maximum average bandwidth a thread saw.
1180 Shortest runtime of threads in the group.
1183 Longest runtime of threads in the group.
1187 Finally, disk statistics are printed with reads first:
1192 Number of I/Os performed by all groups.
1195 Number of merges in the I/O scheduler.
1198 Number of ticks we kept the disk busy.
1201 Total time spent in the disk queue.
1208 It is also possible to get fio to dump the current output while it is
1209 running, without terminating the job. To do that, send fio the \fBUSR1\fR
1212 If the \fB\-\-minimal\fR option is given, the results will be printed in a
1213 semicolon-delimited format suitable for scripted use - a job description
1214 (if provided) follows on a new line. Note that the first
1215 number in the line is the version number. If the output has to be changed
1216 for some reason, this number will be incremented by 1 to signify that
1217 change. The fields are:
1220 .B terse version, fio version, jobname, groupid, error
1224 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1228 .B min, max, mean, standard deviation
1232 .B min, max, mean, standard deviation
1234 Completion latency percentiles (20 fields):
1236 .B Xth percentile=usec
1240 .B min, max, mean, standard deviation
1244 .B min, max, aggregate percentage of total, mean, standard deviation
1250 .B Total I/O \fR(KB)\fP, bandwidth \fR(KB/s)\fP, IOPS, runtime \fR(ms)\fP
1254 .B min, max, mean, standard deviation
1258 .B min, max, mean, standard deviation
1260 Completion latency percentiles (20 fields):
1262 .B Xth percentile=usec
1266 .B min, max, mean, standard deviation
1270 .B min, max, aggregate percentage of total, mean, standard deviation
1276 .B user, system, context switches, major page faults, minor page faults
1279 IO depth distribution:
1281 .B <=1, 2, 4, 8, 16, 32, >=64
1284 IO latency distribution:
1288 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000
1292 .B <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000
1296 Disk utilization (1 for each disk used):
1298 .B name, read ios, write ios, read merges, write merges, read ticks, write ticks, read in-queue time, write in-queue time, disk utilization percentage
1301 Error Info (dependent on continue_on_error, default off):
1303 .B total # errors, first error code
1306 .B text description (if provided in config - appears on newline)
1309 Normally you would run fio as a stand-alone application on the machine
1310 where the IO workload should be generated. However, it is also possible to
1311 run the frontend and backend of fio separately. This makes it possible to
1312 have a fio server running on the machine(s) where the IO workload should
1313 be running, while controlling it from another machine.
1315 To start the server, you would do:
1317 \fBfio \-\-server=args\fR
1319 on that machine, where args defines what fio listens to. The arguments
1320 are of the form 'type:hostname or IP:port'. 'type' is either 'ip' (or ip4)
1321 for TCP/IP v4, 'ip6' for TCP/IP v6, or 'sock' for a local unix domain socket.
1322 'hostname' is either a hostname or IP address, and 'port' is the port to
1323 listen to (only valid for TCP/IP, not a local socket). Some examples:
1327 Start a fio server, listening on all interfaces on the default port (8765).
1329 2) fio --server=ip:hostname,4444
1331 Start a fio server, listening on IP belonging to hostname and on port 4444.
1333 3) fio --server=ip6:::1,4444
1335 Start a fio server, listening on IPv6 localhost ::1 and on port 4444.
1337 4) fio --server=,4444
1339 Start a fio server, listening on all interfaces on port 4444.
1341 5) fio --server=1.2.3.4
1343 Start a fio server, listening on IP 1.2.3.4 on the default port.
1345 6) fio --server=sock:/tmp/fio.sock
1347 Start a fio server, listening on the local socket /tmp/fio.sock.
1349 When a server is running, you can connect to it from a client. The client
1352 fio --local-args --client=server --remote-args <job file(s)>
1354 where --local-args are arguments that are local to the client where it is
1355 running, 'server' is the connect string, and --remote-args and <job file(s)>
1356 are sent to the server. The 'server' string follows the same format as it
1357 does on the server side, to allow IP/hostname/socket and port strings.
1358 You can connect to multiple clients as well, to do that you could run:
1360 fio --client=server2 --client=server2 <job file(s)>
1364 was written by Jens Axboe <jens.axboe@oracle.com>,
1365 now Jens Axboe <jaxboe@fusionio.com>.
1367 This man page was written by Aaron Carroll <aaronc@cse.unsw.edu.au> based
1368 on documentation by Jens Axboe.
1369 .SH "REPORTING BUGS"
1370 Report bugs to the \fBfio\fR mailing list <fio@vger.kernel.org>.
1373 For further documentation see \fBHOWTO\fR and \fBREADME\fR.
1375 Sample jobfiles are available in the \fBexamples\fR directory.