Commit | Line | Data |
---|---|---|
71bfa161 JA |
1 | Table of contents |
2 | ----------------- | |
3 | ||
4 | 1. Overview | |
5 | 2. How fio works | |
6 | 3. Running fio | |
7 | 4. Job file format | |
8 | 5. Detailed list of parameters | |
9 | 6. Normal output | |
10 | 7. Terse output | |
11 | ||
12 | ||
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. | |
21 | ||
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. | |
28 | ||
29 | 2.0 How fio works | |
30 | ----------------- | |
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: | |
38 | ||
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. | |
43 | ||
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 | |
46 | block sizes. | |
47 | ||
48 | IO size How much data are we going to be reading/writing. | |
49 | ||
50 | IO engine How do we issue io? We could be memory mapping the | |
51 | file, we could be using regular read/write, we | |
d0ff85df | 52 | could be using splice, async io, syslet, or even |
71bfa161 JA |
53 | SG (SCSI generic sg). |
54 | ||
6c219763 | 55 | IO depth If the io engine is async, how large a queuing |
71bfa161 JA |
56 | depth do we want to maintain? |
57 | ||
58 | IO type Should we be doing buffered io, or direct/raw io? | |
59 | ||
60 | Num files How many files are we spreading the workload over. | |
61 | ||
62 | Num threads How many threads or processes should we spread | |
63 | this workload over. | |
64 | ||
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 | |
67 | job behaves. | |
68 | ||
69 | ||
70 | 3.0 Running fio | |
71 | --------------- | |
72 | See the README file for command line parameters, there are only a few | |
73 | of them. | |
74 | ||
75 | Running fio is normally the easiest part - you just give it the job file | |
76 | (or job files) as parameters: | |
77 | ||
78 | $ fio job_file | |
79 | ||
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. | |
84 | ||
b4692828 JA |
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 | |
c2b1e753 JA |
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. | |
b4692828 | 96 | |
71bfa161 JA |
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, | |
6c219763 | 99 | such as memory locking, io scheduler switching, and decreasing the nice value. |
71bfa161 JA |
100 | |
101 | ||
102 | 4.0 Job file format | |
103 | ------------------- | |
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 | |
65db0851 JA |
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. | |
71bfa161 | 113 | |
3c54bc46 | 114 | So let's look at a really simple job file that defines two processes, each |
b22989b9 | 115 | randomly reading from a 128MB file. |
71bfa161 JA |
116 | |
117 | ; -- start job file -- | |
118 | [global] | |
119 | rw=randread | |
120 | size=128m | |
121 | ||
122 | [job1] | |
123 | ||
124 | [job2] | |
125 | ||
126 | ; -- end job file -- | |
127 | ||
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 | |
c2b1e753 JA |
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: | |
132 | ||
133 | $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2 | |
134 | ||
71bfa161 | 135 | |
3c54bc46 | 136 | Let's look at an example that has a number of processes writing randomly |
71bfa161 JA |
137 | to files. |
138 | ||
139 | ; -- start job file -- | |
140 | [random-writers] | |
141 | ioengine=libaio | |
142 | iodepth=4 | |
143 | rw=randwrite | |
144 | bs=32k | |
145 | direct=0 | |
146 | size=64m | |
147 | numjobs=4 | |
148 | ||
149 | ; -- end job file -- | |
150 | ||
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 | |
b22989b9 | 153 | increased the buffer size used to 32KB and define numjobs to 4 to |
71bfa161 | 154 | fork 4 identical jobs. The result is 4 processes each randomly writing |
b22989b9 | 155 | to their own 64MB file. Instead of using the above job file, you could |
b4692828 JA |
156 | have given the parameters on the command line. For this case, you would |
157 | specify: | |
158 | ||
159 | $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4 | |
71bfa161 | 160 | |
74929ac2 JA |
161 | 4.1 Environment variables |
162 | ------------------------- | |
163 | ||
3c54bc46 AC |
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 | |
169 | substituted. | |
170 | ||
171 | As an example, let's look at a sample fio invocation and job file: | |
172 | ||
173 | $ SIZE=64m NUMJOBS=4 fio jobfile.fio | |
174 | ||
175 | ; -- start job file -- | |
176 | [random-writers] | |
177 | rw=randwrite | |
178 | size=${SIZE} | |
179 | numjobs=${NUMJOBS} | |
180 | ; -- end job file -- | |
181 | ||
182 | This will expand to the following equivalent job file at runtime: | |
183 | ||
184 | ; -- start job file -- | |
185 | [random-writers] | |
186 | rw=randwrite | |
187 | size=64m | |
188 | numjobs=4 | |
189 | ; -- end job file -- | |
190 | ||
71bfa161 JA |
191 | fio ships with a few example job files, you can also look there for |
192 | inspiration. | |
193 | ||
74929ac2 JA |
194 | 4.2 Reserved keywords |
195 | --------------------- | |
196 | ||
197 | Additionally, fio has a set of reserved keywords that will be replaced | |
198 | internally with the appropriate value. Those keywords are: | |
199 | ||
200 | $pagesize The architecture page size of the running system | |
201 | $mb_memory Megabytes of total memory in the system | |
202 | $ncpus Number of online available CPUs | |
203 | ||
204 | These can be used on the command line or in the job file, and will be | |
205 | automatically substituted with the current system values when the job | |
892a6ffc JA |
206 | is run. Simple math is also supported on these keywords, so you can |
207 | perform actions like: | |
208 | ||
209 | size=8*$mb_memory | |
210 | ||
211 | and get that properly expanded to 8 times the size of memory in the | |
212 | machine. | |
74929ac2 | 213 | |
71bfa161 JA |
214 | |
215 | 5.0 Detailed list of parameters | |
216 | ------------------------------- | |
217 | ||
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: | |
221 | ||
222 | str String. This is a sequence of alpha characters. | |
b09da8fa | 223 | time Integer with possible time suffix. In seconds unless otherwise |
e417fd66 JA |
224 | specified, use eg 10m for 10 minutes. Accepts s/m/h for seconds, |
225 | minutes, and hours. | |
b09da8fa JA |
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 | |
57fc29fa JA |
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 | |
b09da8fa | 231 | out '4096' or just give 4k. The suffixes signify base 2 values, so |
57fc29fa JA |
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. | |
71bfa161 JA |
240 | bool Boolean. Usually parsed as an integer, however only defined for |
241 | true and false (1 and 0). | |
b09da8fa | 242 | irange Integer range with suffix. Allows value range to be given, such |
bf9a3edb | 243 | as 1024-4096. A colon may also be used as the separator, eg |
0c9baf91 JA |
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 | |
f7fa2653 | 246 | int. |
71bfa161 JA |
247 | |
248 | With the above in mind, here follows the complete list of fio job | |
249 | parameters. | |
250 | ||
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 | |
c2b1e753 | 253 | name is used. On the command line this parameter has the |
6c219763 | 254 | special purpose of also signaling the start of a new |
c2b1e753 | 255 | job. |
71bfa161 | 256 | |
61697c37 JA |
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 | |
259 | not parsed. | |
260 | ||
3776041e | 261 | directory=str Prefix filenames with this directory. Used to place files |
71bfa161 JA |
262 | in a different location than "./". |
263 | ||
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 | |
ed92ac0c | 267 | a filename for each of them to override the default. If |
414c2a3e JA |
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 | |
03e20d68 | 274 | filename=/dev/sda:/dev/sdb. On Windows, disk devices are accessed |
ecc314ba BC |
275 | as \\.\PhysicalDrive0 for the first device, \\.\PhysicalDrive1 |
276 | for the second etc. If the wanted filename does need to | |
277 | include a colon, then escape that with a '\' character. | |
278 | For instance, if the filename is "/dev/dsk/foo@3,0:c", | |
279 | then you would use filename="/dev/dsk/foo@3,0\:c". | |
03e20d68 BC |
280 | '-' is a reserved name, meaning stdin or stdout. Which of the |
281 | two depends on the read/write direction set. | |
71bfa161 | 282 | |
bbf6b540 JA |
283 | opendir=str Tell fio to recursively add any file it can find in this |
284 | directory and down the file system tree. | |
285 | ||
3776041e | 286 | lockfile=str Fio defaults to not locking any files before it does |
4d4e80f2 JA |
287 | IO to them. If a file or file descriptor is shared, fio |
288 | can serialize IO to that file to make the end result | |
289 | consistent. This is usual for emulating real workloads that | |
290 | share files. The lock modes are: | |
291 | ||
292 | none No locking. The default. | |
293 | exclusive Only one thread/process may do IO, | |
294 | excluding all others. | |
295 | readwrite Read-write locking on the file. Many | |
296 | readers may access the file at the | |
297 | same time, but writes get exclusive | |
298 | access. | |
299 | ||
300 | The option may be post-fixed with a lock batch number. If | |
301 | set, then each thread/process may do that amount of IOs to | |
bf9a3edb | 302 | the file before giving up the lock. Since lock acquisition is |
4d4e80f2 | 303 | expensive, batching the lock/unlocks will speed up IO. |
29c1349f | 304 | |
d3aad8f2 | 305 | readwrite=str |
71bfa161 JA |
306 | rw=str Type of io pattern. Accepted values are: |
307 | ||
308 | read Sequential reads | |
309 | write Sequential writes | |
310 | randwrite Random writes | |
311 | randread Random reads | |
312 | rw Sequential mixed reads and writes | |
313 | randrw Random mixed reads and writes | |
314 | ||
315 | For the mixed io types, the default is to split them 50/50. | |
316 | For certain types of io the result may still be skewed a bit, | |
211097b2 | 317 | since the speed may be different. It is possible to specify |
38dad62d JA |
318 | a number of IO's to do before getting a new offset, this is |
319 | one by appending a ':<nr>' to the end of the string given. | |
320 | For a random read, it would look like 'rw=randread:8' for | |
321 | passing in an offset modifier with a value of 8. See the | |
322 | 'rw_sequencer' option. | |
323 | ||
324 | rw_sequencer=str If an offset modifier is given by appending a number to | |
325 | the rw=<str> line, then this option controls how that | |
326 | number modifies the IO offset being generated. Accepted | |
327 | values are: | |
328 | ||
329 | sequential Generate sequential offset | |
330 | identical Generate the same offset | |
331 | ||
332 | 'sequential' is only useful for random IO, where fio would | |
333 | normally generate a new random offset for every IO. If you | |
334 | append eg 8 to randread, you would get a new random offset for | |
211097b2 JA |
335 | every 8 IO's. The result would be a seek for only every 8 |
336 | IO's, instead of for every IO. Use rw=randread:8 to specify | |
38dad62d JA |
337 | that. As sequential IO is already sequential, setting |
338 | 'sequential' for that would not result in any differences. | |
339 | 'identical' behaves in a similar fashion, except it sends | |
340 | the same offset 8 number of times before generating a new | |
341 | offset. | |
71bfa161 | 342 | |
90fef2d1 JA |
343 | kb_base=int The base unit for a kilobyte. The defacto base is 2^10, 1024. |
344 | Storage manufacturers like to use 10^3 or 1000 as a base | |
345 | ten unit instead, for obvious reasons. Allow values are | |
346 | 1024 or 1000, with 1024 being the default. | |
347 | ||
ee738499 JA |
348 | randrepeat=bool For random IO workloads, seed the generator in a predictable |
349 | way so that results are repeatable across repetitions. | |
350 | ||
2615cc4b JA |
351 | use_os_rand=bool Fio can either use the random generator supplied by the OS |
352 | to generator random offsets, or it can use it's own internal | |
353 | generator (based on Tausworthe). Default is to use the | |
354 | internal generator, which is often of better quality and | |
355 | faster. | |
356 | ||
7bc8c2cf JA |
357 | fallocate=bool By default, fio will use fallocate() to advise the system |
358 | of the size of the file we are going to write. This can be | |
359 | turned off with fallocate=0. May not be available on all | |
9b836561 BC |
360 | supported platforms. If using ZFS on Solaris this must be |
361 | set to 0 because ZFS doesn't support it. | |
7bc8c2cf | 362 | |
d2f3ac35 JA |
363 | fadvise_hint=bool By default, fio will use fadvise() to advise the kernel |
364 | on what IO patterns it is likely to issue. Sometimes you | |
365 | want to test specific IO patterns without telling the | |
366 | kernel about it, in which case you can disable this option. | |
367 | If set, fio will use POSIX_FADV_SEQUENTIAL for sequential | |
368 | IO and POSIX_FADV_RANDOM for random IO. | |
369 | ||
f7fa2653 | 370 | size=int The total size of file io for this job. Fio will run until |
7616cafe JA |
371 | this many bytes has been transferred, unless runtime is |
372 | limited by other options (such as 'runtime', for instance). | |
3776041e | 373 | Unless specific nrfiles and filesize options are given, |
7616cafe | 374 | fio will divide this size between the available files |
d6667268 JA |
375 | specified by the job. If not set, fio will use the full |
376 | size of the given files or devices. If the the files | |
377 | do not exist, size must be given. | |
71bfa161 | 378 | |
f7fa2653 | 379 | filesize=int Individual file sizes. May be a range, in which case fio |
9c60ce64 JA |
380 | will select sizes for files at random within the given range |
381 | and limited to 'size' in total (if that is given). If not | |
382 | given, each created file is the same size. | |
383 | ||
74586c1e JA |
384 | fill_device=bool |
385 | fill_fs=bool Sets size to something really large and waits for ENOSPC (no | |
aa31f1f1 | 386 | space left on device) as the terminating condition. Only makes |
3ce9dcaf | 387 | sense with sequential write. For a read workload, the mount |
4f12432e JA |
388 | point will be filled first then IO started on the result. This |
389 | option doesn't make sense if operating on a raw device node, | |
390 | since the size of that is already known by the file system. | |
391 | Additionally, writing beyond end-of-device will not return | |
392 | ENOSPC there. | |
aa31f1f1 | 393 | |
f7fa2653 JA |
394 | blocksize=int |
395 | bs=int The block size used for the io units. Defaults to 4k. Values | |
396 | can be given for both read and writes. If a single int is | |
397 | given, it will apply to both. If a second int is specified | |
f90eff5a JA |
398 | after a comma, it will apply to writes only. In other words, |
399 | the format is either bs=read_and_write or bs=read,write. | |
400 | bs=4k,8k will thus use 4k blocks for reads, and 8k blocks | |
787f7e95 JA |
401 | for writes. If you only wish to set the write size, you |
402 | can do so by passing an empty read size - bs=,8k will set | |
403 | 8k for writes and leave the read default value. | |
a00735e6 | 404 | |
2b7a01d0 JA |
405 | blockalign=int |
406 | ba=int At what boundary to align random IO offsets. Defaults to | |
407 | the same as 'blocksize' the minimum blocksize given. | |
408 | Minimum alignment is typically 512b for using direct IO, | |
409 | though it usually depends on the hardware block size. This | |
410 | option is mutually exclusive with using a random map for | |
411 | files, so it will turn off that option. | |
412 | ||
d3aad8f2 | 413 | blocksize_range=irange |
71bfa161 JA |
414 | bsrange=irange Instead of giving a single block size, specify a range |
415 | and fio will mix the issued io block sizes. The issued | |
416 | io unit will always be a multiple of the minimum value | |
f90eff5a JA |
417 | given (also see bs_unaligned). Applies to both reads and |
418 | writes, however a second range can be given after a comma. | |
419 | See bs=. | |
a00735e6 | 420 | |
564ca972 JA |
421 | bssplit=str Sometimes you want even finer grained control of the |
422 | block sizes issued, not just an even split between them. | |
423 | This option allows you to weight various block sizes, | |
424 | so that you are able to define a specific amount of | |
425 | block sizes issued. The format for this option is: | |
426 | ||
427 | bssplit=blocksize/percentage:blocksize/percentage | |
428 | ||
429 | for as many block sizes as needed. So if you want to define | |
430 | a workload that has 50% 64k blocks, 10% 4k blocks, and | |
431 | 40% 32k blocks, you would write: | |
432 | ||
433 | bssplit=4k/10:64k/50:32k/40 | |
434 | ||
435 | Ordering does not matter. If the percentage is left blank, | |
436 | fio will fill in the remaining values evenly. So a bssplit | |
437 | option like this one: | |
438 | ||
439 | bssplit=4k/50:1k/:32k/ | |
440 | ||
441 | would have 50% 4k ios, and 25% 1k and 32k ios. The percentages | |
442 | always add up to 100, if bssplit is given a range that adds | |
443 | up to more, it will error out. | |
444 | ||
720e84ad JA |
445 | bssplit also supports giving separate splits to reads and |
446 | writes. The format is identical to what bs= accepts. You | |
447 | have to separate the read and write parts with a comma. So | |
448 | if you want a workload that has 50% 2k reads and 50% 4k reads, | |
449 | while having 90% 4k writes and 10% 8k writes, you would | |
450 | specify: | |
451 | ||
452 | bssplit=2k/50:4k/50,4k/90,8k/10 | |
453 | ||
d3aad8f2 | 454 | blocksize_unaligned |
690adba3 JA |
455 | bs_unaligned If this option is given, any byte size value within bsrange |
456 | may be used as a block range. This typically wont work with | |
457 | direct IO, as that normally requires sector alignment. | |
71bfa161 | 458 | |
e9459e5a JA |
459 | zero_buffers If this option is given, fio will init the IO buffers to |
460 | all zeroes. The default is to fill them with random data. | |
461 | ||
5973cafb JA |
462 | refill_buffers If this option is given, fio will refill the IO buffers |
463 | on every submit. The default is to only fill it at init | |
464 | time and reuse that data. Only makes sense if zero_buffers | |
41ccd845 JA |
465 | isn't specified, naturally. If data verification is enabled, |
466 | refill_buffers is also automatically enabled. | |
5973cafb | 467 | |
71bfa161 JA |
468 | nrfiles=int Number of files to use for this job. Defaults to 1. |
469 | ||
390b1537 JA |
470 | openfiles=int Number of files to keep open at the same time. Defaults to |
471 | the same as nrfiles, can be set smaller to limit the number | |
472 | simultaneous opens. | |
473 | ||
5af1c6f3 JA |
474 | file_service_type=str Defines how fio decides which file from a job to |
475 | service next. The following types are defined: | |
476 | ||
477 | random Just choose a file at random. | |
478 | ||
479 | roundrobin Round robin over open files. This | |
480 | is the default. | |
481 | ||
a086c257 JA |
482 | sequential Finish one file before moving on to |
483 | the next. Multiple files can still be | |
484 | open depending on 'openfiles'. | |
485 | ||
1907dbc6 JA |
486 | The string can have a number appended, indicating how |
487 | often to switch to a new file. So if option random:4 is | |
488 | given, fio will switch to a new random file after 4 ios | |
489 | have been issued. | |
490 | ||
71bfa161 JA |
491 | ioengine=str Defines how the job issues io to the file. The following |
492 | types are defined: | |
493 | ||
494 | sync Basic read(2) or write(2) io. lseek(2) is | |
495 | used to position the io location. | |
496 | ||
a31041ea | 497 | psync Basic pread(2) or pwrite(2) io. |
498 | ||
e05af9e5 | 499 | vsync Basic readv(2) or writev(2) IO. |
1d2af02a | 500 | |
15d182aa JA |
501 | libaio Linux native asynchronous io. Note that Linux |
502 | may only support queued behaviour with | |
503 | non-buffered IO (set direct=1 or buffered=0). | |
71bfa161 JA |
504 | |
505 | posixaio glibc posix asynchronous io. | |
506 | ||
417f0068 JA |
507 | solarisaio Solaris native asynchronous io. |
508 | ||
03e20d68 BC |
509 | windowsaio Windows native asynchronous io. |
510 | ||
71bfa161 JA |
511 | mmap File is memory mapped and data copied |
512 | to/from using memcpy(3). | |
513 | ||
514 | splice splice(2) is used to transfer the data and | |
515 | vmsplice(2) to transfer data from user | |
516 | space to the kernel. | |
517 | ||
d0ff85df JA |
518 | syslet-rw Use the syslet system calls to make |
519 | regular read/write async. | |
520 | ||
71bfa161 | 521 | sg SCSI generic sg v3 io. May either be |
6c219763 | 522 | synchronous using the SG_IO ioctl, or if |
71bfa161 JA |
523 | the target is an sg character device |
524 | we use read(2) and write(2) for asynchronous | |
525 | io. | |
526 | ||
a94ea28b JA |
527 | null Doesn't transfer any data, just pretends |
528 | to. This is mainly used to exercise fio | |
529 | itself and for debugging/testing purposes. | |
530 | ||
ed92ac0c JA |
531 | net Transfer over the network to given host:port. |
532 | 'filename' must be set appropriately to | |
414c2a3e | 533 | filename=host/port/protocol regardless of send |
ed92ac0c | 534 | or receive, if the latter only the port |
414c2a3e JA |
535 | argument is used. 'host' may be an IP address |
536 | or hostname, port is the port number to be used, | |
537 | and protocol may be 'udp' or 'tcp'. If no | |
538 | protocol is given, TCP is used. | |
ed92ac0c | 539 | |
9cce02e8 JA |
540 | netsplice Like net, but uses splice/vmsplice to |
541 | map data and send/receive. | |
542 | ||
53aec0a4 | 543 | cpuio Doesn't transfer any data, but burns CPU |
ba0fbe10 JA |
544 | cycles according to the cpuload= and |
545 | cpucycle= options. Setting cpuload=85 | |
546 | will cause that job to do nothing but burn | |
36ecec83 GP |
547 | 85% of the CPU. In case of SMP machines, |
548 | use numjobs=<no_of_cpu> to get desired CPU | |
549 | usage, as the cpuload only loads a single | |
550 | CPU at the desired rate. | |
ba0fbe10 | 551 | |
e9a1806f JA |
552 | guasi The GUASI IO engine is the Generic Userspace |
553 | Asyncronous Syscall Interface approach | |
554 | to async IO. See | |
555 | ||
556 | http://www.xmailserver.org/guasi-lib.html | |
557 | ||
558 | for more info on GUASI. | |
559 | ||
8a7bd877 JA |
560 | external Prefix to specify loading an external |
561 | IO engine object file. Append the engine | |
562 | filename, eg ioengine=external:/tmp/foo.o | |
563 | to load ioengine foo.o in /tmp. | |
564 | ||
71bfa161 JA |
565 | iodepth=int This defines how many io units to keep in flight against |
566 | the file. The default is 1 for each file defined in this | |
567 | job, can be overridden with a larger value for higher | |
ee72ca09 JA |
568 | concurrency. Note that increasing iodepth beyond 1 will not |
569 | affect synchronous ioengines (except for small degress when | |
9b836561 | 570 | verify_async is in use). Even async engines may impose OS |
ee72ca09 JA |
571 | restrictions causing the desired depth not to be achieved. |
572 | This may happen on Linux when using libaio and not setting | |
573 | direct=1, since buffered IO is not async on that OS. Keep an | |
574 | eye on the IO depth distribution in the fio output to verify | |
575 | that the achieved depth is as expected. Default: 1. | |
71bfa161 | 576 | |
4950421a | 577 | iodepth_batch_submit=int |
cb5ab512 | 578 | iodepth_batch=int This defines how many pieces of IO to submit at once. |
89e820f6 JA |
579 | It defaults to 1 which means that we submit each IO |
580 | as soon as it is available, but can be raised to submit | |
581 | bigger batches of IO at the time. | |
cb5ab512 | 582 | |
4950421a JA |
583 | iodepth_batch_complete=int This defines how many pieces of IO to retrieve |
584 | at once. It defaults to 1 which means that we'll ask | |
585 | for a minimum of 1 IO in the retrieval process from | |
586 | the kernel. The IO retrieval will go on until we | |
587 | hit the limit set by iodepth_low. If this variable is | |
588 | set to 0, then fio will always check for completed | |
589 | events before queuing more IO. This helps reduce | |
590 | IO latency, at the cost of more retrieval system calls. | |
591 | ||
e916b390 JA |
592 | iodepth_low=int The low water mark indicating when to start filling |
593 | the queue again. Defaults to the same as iodepth, meaning | |
594 | that fio will attempt to keep the queue full at all times. | |
595 | If iodepth is set to eg 16 and iodepth_low is set to 4, then | |
596 | after fio has filled the queue of 16 requests, it will let | |
597 | the depth drain down to 4 before starting to fill it again. | |
598 | ||
71bfa161 | 599 | direct=bool If value is true, use non-buffered io. This is usually |
9b836561 | 600 | O_DIRECT. Note that ZFS on Solaris doesn't support direct io. |
76a43db4 JA |
601 | |
602 | buffered=bool If value is true, use buffered io. This is the opposite | |
603 | of the 'direct' option. Defaults to true. | |
71bfa161 | 604 | |
f7fa2653 | 605 | offset=int Start io at the given offset in the file. The data before |
71bfa161 JA |
606 | the given offset will not be touched. This effectively |
607 | caps the file size at real_size - offset. | |
608 | ||
609 | fsync=int If writing to a file, issue a sync of the dirty data | |
610 | for every number of blocks given. For example, if you give | |
611 | 32 as a parameter, fio will sync the file for every 32 | |
612 | writes issued. If fio is using non-buffered io, we may | |
613 | not sync the file. The exception is the sg io engine, which | |
6c219763 | 614 | synchronizes the disk cache anyway. |
71bfa161 | 615 | |
e76b1da4 | 616 | fdatasync=int Like fsync= but uses fdatasync() to only sync data and not |
5f9099ea | 617 | metadata blocks. |
e72fa4d4 JA |
618 | In FreeBSD there is no fdatasync(), this falls back to |
619 | using fsync() | |
5f9099ea | 620 | |
e76b1da4 JA |
621 | sync_file_range=str:val Use sync_file_range() for every 'val' number of |
622 | write operations. Fio will track range of writes that | |
623 | have happened since the last sync_file_range() call. 'str' | |
624 | can currently be one or more of: | |
625 | ||
626 | wait_before SYNC_FILE_RANGE_WAIT_BEFORE | |
627 | write SYNC_FILE_RANGE_WRITE | |
628 | wait_after SYNC_FILE_RANGE_WAIT_AFTER | |
629 | ||
630 | So if you do sync_file_range=wait_before,write:8, fio would | |
631 | use SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE for | |
632 | every 8 writes. Also see the sync_file_range(2) man page. | |
633 | This option is Linux specific. | |
634 | ||
5036fc1e JA |
635 | overwrite=bool If true, writes to a file will always overwrite existing |
636 | data. If the file doesn't already exist, it will be | |
637 | created before the write phase begins. If the file exists | |
638 | and is large enough for the specified write phase, nothing | |
639 | will be done. | |
71bfa161 JA |
640 | |
641 | end_fsync=bool If true, fsync file contents when the job exits. | |
642 | ||
ebb1415f JA |
643 | fsync_on_close=bool If true, fio will fsync() a dirty file on close. |
644 | This differs from end_fsync in that it will happen on every | |
645 | file close, not just at the end of the job. | |
646 | ||
71bfa161 JA |
647 | rwmixread=int How large a percentage of the mix should be reads. |
648 | ||
649 | rwmixwrite=int How large a percentage of the mix should be writes. If both | |
650 | rwmixread and rwmixwrite is given and the values do not add | |
651 | up to 100%, the latter of the two will be used to override | |
c35dd7a6 JA |
652 | the first. This may interfere with a given rate setting, |
653 | if fio is asked to limit reads or writes to a certain rate. | |
654 | If that is the case, then the distribution may be skewed. | |
71bfa161 | 655 | |
bb8895e0 JA |
656 | norandommap Normally fio will cover every block of the file when doing |
657 | random IO. If this option is given, fio will just get a | |
658 | new random offset without looking at past io history. This | |
659 | means that some blocks may not be read or written, and that | |
660 | some blocks may be read/written more than once. This option | |
8347239a JA |
661 | is mutually exclusive with verify= if and only if multiple |
662 | blocksizes (via bsrange=) are used, since fio only tracks | |
663 | complete rewrites of blocks. | |
bb8895e0 | 664 | |
2b386d25 JA |
665 | softrandommap See norandommap. If fio runs with the random block map enabled |
666 | and it fails to allocate the map, if this option is set it | |
667 | will continue without a random block map. As coverage will | |
668 | not be as complete as with random maps, this option is | |
669 | disabled by default. | |
670 | ||
71bfa161 JA |
671 | nice=int Run the job with the given nice value. See man nice(2). |
672 | ||
673 | prio=int Set the io priority value of this job. Linux limits us to | |
674 | a positive value between 0 and 7, with 0 being the highest. | |
675 | See man ionice(1). | |
676 | ||
677 | prioclass=int Set the io priority class. See man ionice(1). | |
678 | ||
679 | thinktime=int Stall the job x microseconds after an io has completed before | |
680 | issuing the next. May be used to simulate processing being | |
48097d5c JA |
681 | done by an application. See thinktime_blocks and |
682 | thinktime_spin. | |
683 | ||
684 | thinktime_spin=int | |
685 | Only valid if thinktime is set - pretend to spend CPU time | |
686 | doing something with the data received, before falling back | |
687 | to sleeping for the rest of the period specified by | |
688 | thinktime. | |
9c1f7434 JA |
689 | |
690 | thinktime_blocks | |
691 | Only valid if thinktime is set - control how many blocks | |
692 | to issue, before waiting 'thinktime' usecs. If not set, | |
693 | defaults to 1 which will make fio wait 'thinktime' usecs | |
694 | after every block. | |
71bfa161 | 695 | |
581e7141 | 696 | rate=int Cap the bandwidth used by this job. The number is in bytes/sec, |
b09da8fa | 697 | the normal suffix rules apply. You can use rate=500k to limit |
581e7141 JA |
698 | reads and writes to 500k each, or you can specify read and |
699 | writes separately. Using rate=1m,500k would limit reads to | |
700 | 1MB/sec and writes to 500KB/sec. Capping only reads or | |
701 | writes can be done with rate=,500k or rate=500k,. The former | |
702 | will only limit writes (to 500KB/sec), the latter will only | |
703 | limit reads. | |
71bfa161 JA |
704 | |
705 | ratemin=int Tell fio to do whatever it can to maintain at least this | |
4e991c23 | 706 | bandwidth. Failing to meet this requirement, will cause |
581e7141 JA |
707 | the job to exit. The same format as rate is used for |
708 | read vs write separation. | |
4e991c23 JA |
709 | |
710 | rate_iops=int Cap the bandwidth to this number of IOPS. Basically the same | |
711 | as rate, just specified independently of bandwidth. If the | |
712 | job is given a block size range instead of a fixed value, | |
581e7141 JA |
713 | the smallest block size is used as the metric. The same format |
714 | as rate is used for read vs write seperation. | |
4e991c23 JA |
715 | |
716 | rate_iops_min=int If fio doesn't meet this rate of IO, it will cause | |
581e7141 JA |
717 | the job to exit. The same format as rate is used for read vs |
718 | write seperation. | |
71bfa161 JA |
719 | |
720 | ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number | |
6c219763 | 721 | of milliseconds. |
71bfa161 JA |
722 | |
723 | cpumask=int Set the CPU affinity of this job. The parameter given is a | |
a08bc17f JA |
724 | bitmask of allowed CPU's the job may run on. So if you want |
725 | the allowed CPUs to be 1 and 5, you would pass the decimal | |
726 | value of (1 << 1 | 1 << 5), or 34. See man | |
7dbb6eba | 727 | sched_setaffinity(2). This may not work on all supported |
b0ea08ce JA |
728 | operating systems or kernel versions. This option doesn't |
729 | work well for a higher CPU count than what you can store in | |
730 | an integer mask, so it can only control cpus 1-32. For | |
731 | boxes with larger CPU counts, use cpus_allowed. | |
71bfa161 | 732 | |
d2e268b0 JA |
733 | cpus_allowed=str Controls the same options as cpumask, but it allows a text |
734 | setting of the permitted CPUs instead. So to use CPUs 1 and | |
62a7273d JA |
735 | 5, you would specify cpus_allowed=1,5. This options also |
736 | allows a range of CPUs. Say you wanted a binding to CPUs | |
737 | 1, 5, and 8-15, you would set cpus_allowed=1,5,8-15. | |
d2e268b0 | 738 | |
e417fd66 | 739 | startdelay=time Start this job the specified number of seconds after fio |
71bfa161 JA |
740 | has started. Only useful if the job file contains several |
741 | jobs, and you want to delay starting some jobs to a certain | |
742 | time. | |
743 | ||
e417fd66 | 744 | runtime=time Tell fio to terminate processing after the specified number |
71bfa161 JA |
745 | of seconds. It can be quite hard to determine for how long |
746 | a specified job will run, so this parameter is handy to | |
747 | cap the total runtime to a given time. | |
748 | ||
cf4464ca | 749 | time_based If set, fio will run for the duration of the runtime |
bf9a3edb | 750 | specified even if the file(s) are completely read or |
cf4464ca JA |
751 | written. It will simply loop over the same workload |
752 | as many times as the runtime allows. | |
753 | ||
e417fd66 | 754 | ramp_time=time If set, fio will run the specified workload for this amount |
721938ae JA |
755 | of time before logging any performance numbers. Useful for |
756 | letting performance settle before logging results, thus | |
b29ee5b3 JA |
757 | minimizing the runtime required for stable results. Note |
758 | that the ramp_time is considered lead in time for a job, | |
759 | thus it will increase the total runtime if a special timeout | |
760 | or runtime is specified. | |
721938ae | 761 | |
71bfa161 JA |
762 | invalidate=bool Invalidate the buffer/page cache parts for this file prior |
763 | to starting io. Defaults to true. | |
764 | ||
765 | sync=bool Use sync io for buffered writes. For the majority of the | |
766 | io engines, this means using O_SYNC. | |
767 | ||
d3aad8f2 | 768 | iomem=str |
71bfa161 JA |
769 | mem=str Fio can use various types of memory as the io unit buffer. |
770 | The allowed values are: | |
771 | ||
772 | malloc Use memory from malloc(3) as the buffers. | |
773 | ||
774 | shm Use shared memory as the buffers. Allocated | |
775 | through shmget(2). | |
776 | ||
74b025b0 JA |
777 | shmhuge Same as shm, but use huge pages as backing. |
778 | ||
313cb206 JA |
779 | mmap Use mmap to allocate buffers. May either be |
780 | anonymous memory, or can be file backed if | |
781 | a filename is given after the option. The | |
782 | format is mem=mmap:/path/to/file. | |
71bfa161 | 783 | |
d0bdaf49 JA |
784 | mmaphuge Use a memory mapped huge file as the buffer |
785 | backing. Append filename after mmaphuge, ala | |
786 | mem=mmaphuge:/hugetlbfs/file | |
787 | ||
71bfa161 | 788 | The area allocated is a function of the maximum allowed |
5394ae5f JA |
789 | bs size for the job, multiplied by the io depth given. Note |
790 | that for shmhuge and mmaphuge to work, the system must have | |
791 | free huge pages allocated. This can normally be checked | |
792 | and set by reading/writing /proc/sys/vm/nr_hugepages on a | |
b22989b9 | 793 | Linux system. Fio assumes a huge page is 4MB in size. So |
5394ae5f JA |
794 | to calculate the number of huge pages you need for a given |
795 | job file, add up the io depth of all jobs (normally one unless | |
796 | iodepth= is used) and multiply by the maximum bs set. Then | |
797 | divide that number by the huge page size. You can see the | |
798 | size of the huge pages in /proc/meminfo. If no huge pages | |
799 | are allocated by having a non-zero number in nr_hugepages, | |
56bb17f2 | 800 | using mmaphuge or shmhuge will fail. Also see hugepage-size. |
5394ae5f JA |
801 | |
802 | mmaphuge also needs to have hugetlbfs mounted and the file | |
803 | location should point there. So if it's mounted in /huge, | |
804 | you would use mem=mmaphuge:/huge/somefile. | |
71bfa161 | 805 | |
d529ee19 JA |
806 | iomem_align=int This indiciates the memory alignment of the IO memory buffers. |
807 | Note that the given alignment is applied to the first IO unit | |
808 | buffer, if using iodepth the alignment of the following buffers | |
809 | are given by the bs used. In other words, if using a bs that is | |
810 | a multiple of the page sized in the system, all buffers will | |
811 | be aligned to this value. If using a bs that is not page | |
812 | aligned, the alignment of subsequent IO memory buffers is the | |
813 | sum of the iomem_align and bs used. | |
814 | ||
f7fa2653 | 815 | hugepage-size=int |
56bb17f2 | 816 | Defines the size of a huge page. Must at least be equal |
b22989b9 | 817 | to the system setting, see /proc/meminfo. Defaults to 4MB. |
c51074e7 JA |
818 | Should probably always be a multiple of megabytes, so using |
819 | hugepage-size=Xm is the preferred way to set this to avoid | |
820 | setting a non-pow-2 bad value. | |
56bb17f2 | 821 | |
71bfa161 JA |
822 | exitall When one job finishes, terminate the rest. The default is |
823 | to wait for each job to finish, sometimes that is not the | |
824 | desired action. | |
825 | ||
826 | bwavgtime=int Average the calculated bandwidth over the given time. Value | |
6c219763 | 827 | is specified in milliseconds. |
71bfa161 JA |
828 | |
829 | create_serialize=bool If true, serialize the file creating for the jobs. | |
830 | This may be handy to avoid interleaving of data | |
831 | files, which may greatly depend on the filesystem | |
832 | used and even the number of processors in the system. | |
833 | ||
834 | create_fsync=bool fsync the data file after creation. This is the | |
835 | default. | |
836 | ||
814452bd JA |
837 | create_on_open=bool Don't pre-setup the files for IO, just create open() |
838 | when it's time to do IO to that file. | |
839 | ||
afad68f7 | 840 | pre_read=bool If this is given, files will be pre-read into memory before |
34f1c044 JA |
841 | starting the given IO operation. This will also clear |
842 | the 'invalidate' flag, since it is pointless to pre-read | |
9c0d2241 JA |
843 | and then drop the cache. This will only work for IO engines |
844 | that are seekable, since they allow you to read the same data | |
845 | multiple times. Thus it will not work on eg network or splice | |
846 | IO. | |
afad68f7 | 847 | |
e545a6ce | 848 | unlink=bool Unlink the job files when done. Not the default, as repeated |
bf9a3edb JA |
849 | runs of that job would then waste time recreating the file |
850 | set again and again. | |
71bfa161 JA |
851 | |
852 | loops=int Run the specified number of iterations of this job. Used | |
853 | to repeat the same workload a given number of times. Defaults | |
854 | to 1. | |
855 | ||
68e1f29a | 856 | do_verify=bool Run the verify phase after a write phase. Only makes sense if |
e84c73a8 SL |
857 | verify is set. Defaults to 1. |
858 | ||
71bfa161 JA |
859 | verify=str If writing to a file, fio can verify the file contents |
860 | after each iteration of the job. The allowed values are: | |
861 | ||
862 | md5 Use an md5 sum of the data area and store | |
863 | it in the header of each block. | |
864 | ||
17dc34df JA |
865 | crc64 Use an experimental crc64 sum of the data |
866 | area and store it in the header of each | |
867 | block. | |
868 | ||
bac39e0e JA |
869 | crc32c Use a crc32c sum of the data area and store |
870 | it in the header of each block. | |
871 | ||
3845591f | 872 | crc32c-intel Use hardware assisted crc32c calcuation |
0539d758 JA |
873 | provided on SSE4.2 enabled processors. Falls |
874 | back to regular software crc32c, if not | |
875 | supported by the system. | |
3845591f | 876 | |
71bfa161 JA |
877 | crc32 Use a crc32 sum of the data area and store |
878 | it in the header of each block. | |
879 | ||
969f7ed3 JA |
880 | crc16 Use a crc16 sum of the data area and store |
881 | it in the header of each block. | |
882 | ||
17dc34df JA |
883 | crc7 Use a crc7 sum of the data area and store |
884 | it in the header of each block. | |
885 | ||
cd14cc10 JA |
886 | sha512 Use sha512 as the checksum function. |
887 | ||
888 | sha256 Use sha256 as the checksum function. | |
889 | ||
7c353ceb JA |
890 | sha1 Use optimized sha1 as the checksum function. |
891 | ||
7437ee87 SL |
892 | meta Write extra information about each io |
893 | (timestamp, block number etc.). The block | |
996093bb | 894 | number is verified. See also verify_pattern. |
7437ee87 | 895 | |
36690c9b JA |
896 | null Only pretend to verify. Useful for testing |
897 | internals with ioengine=null, not for much | |
898 | else. | |
899 | ||
6c219763 | 900 | This option can be used for repeated burn-in tests of a |
71bfa161 | 901 | system to make sure that the written data is also |
b892dc08 JA |
902 | correctly read back. If the data direction given is |
903 | a read or random read, fio will assume that it should | |
904 | verify a previously written file. If the data direction | |
905 | includes any form of write, the verify will be of the | |
906 | newly written data. | |
71bfa161 | 907 | |
160b966d JA |
908 | verifysort=bool If set, fio will sort written verify blocks when it deems |
909 | it faster to read them back in a sorted manner. This is | |
910 | often the case when overwriting an existing file, since | |
911 | the blocks are already laid out in the file system. You | |
912 | can ignore this option unless doing huge amounts of really | |
913 | fast IO where the red-black tree sorting CPU time becomes | |
914 | significant. | |
3f9f4e26 | 915 | |
f7fa2653 | 916 | verify_offset=int Swap the verification header with data somewhere else |
546a9142 SL |
917 | in the block before writing. Its swapped back before |
918 | verifying. | |
919 | ||
f7fa2653 | 920 | verify_interval=int Write the verification header at a finer granularity |
3f9f4e26 SL |
921 | than the blocksize. It will be written for chunks the |
922 | size of header_interval. blocksize should divide this | |
923 | evenly. | |
90059d65 | 924 | |
0e92f873 | 925 | verify_pattern=str If set, fio will fill the io buffers with this |
e28218f3 SL |
926 | pattern. Fio defaults to filling with totally random |
927 | bytes, but sometimes it's interesting to fill with a known | |
928 | pattern for io verification purposes. Depending on the | |
929 | width of the pattern, fio will fill 1/2/3/4 bytes of the | |
0e92f873 RR |
930 | buffer at the time(it can be either a decimal or a hex number). |
931 | The verify_pattern if larger than a 32-bit quantity has to | |
996093bb JA |
932 | be a hex number that starts with either "0x" or "0X". Use |
933 | with verify=meta. | |
e28218f3 | 934 | |
68e1f29a | 935 | verify_fatal=bool Normally fio will keep checking the entire contents |
a12a3b4d JA |
936 | before quitting on a block verification failure. If this |
937 | option is set, fio will exit the job on the first observed | |
938 | failure. | |
e8462bd8 | 939 | |
b463e936 JA |
940 | verify_dump=bool If set, dump the contents of both the original data |
941 | block and the data block we read off disk to files. This | |
942 | allows later analysis to inspect just what kind of data | |
943 | corruption occurred. On by default. | |
944 | ||
e8462bd8 JA |
945 | verify_async=int Fio will normally verify IO inline from the submitting |
946 | thread. This option takes an integer describing how many | |
947 | async offload threads to create for IO verification instead, | |
948 | causing fio to offload the duty of verifying IO contents | |
c85c324c JA |
949 | to one or more separate threads. If using this offload |
950 | option, even sync IO engines can benefit from using an | |
951 | iodepth setting higher than 1, as it allows them to have | |
952 | IO in flight while verifies are running. | |
e8462bd8 JA |
953 | |
954 | verify_async_cpus=str Tell fio to set the given CPU affinity on the | |
955 | async IO verification threads. See cpus_allowed for the | |
956 | format used. | |
6f87418f JA |
957 | |
958 | verify_backlog=int Fio will normally verify the written contents of a | |
959 | job that utilizes verify once that job has completed. In | |
960 | other words, everything is written then everything is read | |
961 | back and verified. You may want to verify continually | |
962 | instead for a variety of reasons. Fio stores the meta data | |
963 | associated with an IO block in memory, so for large | |
964 | verify workloads, quite a bit of memory would be used up | |
965 | holding this meta data. If this option is enabled, fio | |
f42195a3 JA |
966 | will write only N blocks before verifying these blocks. |
967 | ||
6f87418f JA |
968 | will verify the previously written blocks before continuing |
969 | to write new ones. | |
970 | ||
971 | verify_backlog_batch=int Control how many blocks fio will verify | |
972 | if verify_backlog is set. If not set, will default to | |
973 | the value of verify_backlog (meaning the entire queue | |
f42195a3 JA |
974 | is read back and verified). If verify_backlog_batch is |
975 | less than verify_backlog then not all blocks will be verified, | |
976 | if verify_backlog_batch is larger than verify_backlog, some | |
977 | blocks will be verified more than once. | |
160b966d | 978 | |
71bfa161 JA |
979 | stonewall Wait for preceeding jobs in the job file to exit, before |
980 | starting this one. Can be used to insert serialization | |
b3d62a75 JA |
981 | points in the job file. A stone wall also implies starting |
982 | a new reporting group. | |
983 | ||
984 | new_group Start a new reporting group. If this option isn't given, | |
985 | jobs in a file will be part of the same reporting group | |
bf9a3edb | 986 | unless separated by a stone wall (or if it's a group |
b3d62a75 | 987 | by itself, with the numjobs option). |
71bfa161 JA |
988 | |
989 | numjobs=int Create the specified number of clones of this job. May be | |
990 | used to setup a larger number of threads/processes doing | |
fa28c85a JA |
991 | the same thing. We regard that grouping of jobs as a |
992 | specific group. | |
993 | ||
994 | group_reporting If 'numjobs' is set, it may be interesting to display | |
995 | statistics for the group as a whole instead of for each | |
996 | individual job. This is especially true of 'numjobs' is | |
997 | large, looking at individual thread/process output quickly | |
998 | becomes unwieldy. If 'group_reporting' is specified, fio | |
999 | will show the final report per-group instead of per-job. | |
71bfa161 JA |
1000 | |
1001 | thread fio defaults to forking jobs, however if this option is | |
1002 | given, fio will use pthread_create(3) to create threads | |
1003 | instead. | |
1004 | ||
f7fa2653 | 1005 | zonesize=int Divide a file into zones of the specified size. See zoneskip. |
71bfa161 | 1006 | |
f7fa2653 | 1007 | zoneskip=int Skip the specified number of bytes when zonesize data has |
71bfa161 JA |
1008 | been read. The two zone options can be used to only do |
1009 | io on zones of a file. | |
1010 | ||
076efc7c | 1011 | write_iolog=str Write the issued io patterns to the specified file. See |
5b42a488 SH |
1012 | read_iolog. Specify a separate file for each job, otherwise |
1013 | the iologs will be interspersed and the file may be corrupt. | |
71bfa161 | 1014 | |
076efc7c | 1015 | read_iolog=str Open an iolog with the specified file name and replay the |
71bfa161 | 1016 | io patterns it contains. This can be used to store a |
6df8adaa JA |
1017 | workload and replay it sometime later. The iolog given |
1018 | may also be a blktrace binary file, which allows fio | |
1019 | to replay a workload captured by blktrace. See blktrace | |
1020 | for how to capture such logging data. For blktrace replay, | |
1021 | the file needs to be turned into a blkparse binary data | |
ea3e51c3 | 1022 | file first (blkparse <device> -o /dev/null -d file_for_fio.bin). |
64bbb865 DN |
1023 | |
1024 | replay_no_stall=int When replaying I/O with read_iolog the default behavior | |
62776229 JA |
1025 | is to attempt to respect the time stamps within the log and |
1026 | replay them with the appropriate delay between IOPS. By | |
1027 | setting this variable fio will not respect the timestamps and | |
1028 | attempt to replay them as fast as possible while still | |
1029 | respecting ordering. The result is the same I/O pattern to a | |
1030 | given device, but different timings. | |
71bfa161 | 1031 | |
d1c46c04 DN |
1032 | replay_redirect=str While replaying I/O patterns using read_iolog the |
1033 | default behavior is to replay the IOPS onto the major/minor | |
1034 | device that each IOP was recorded from. This is sometimes | |
1035 | undesireable because on a different machine those major/minor | |
1036 | numbers can map to a different device. Changing hardware on | |
1037 | the same system can also result in a different major/minor | |
1038 | mapping. Replay_redirect causes all IOPS to be replayed onto | |
1039 | the single specified device regardless of the device it was | |
1040 | recorded from. i.e. replay_redirect=/dev/sdc would cause all | |
1041 | IO in the blktrace to be replayed onto /dev/sdc. This means | |
1042 | multiple devices will be replayed onto a single, if the trace | |
1043 | contains multiple devices. If you want multiple devices to be | |
1044 | replayed concurrently to multiple redirected devices you must | |
1045 | blkparse your trace into separate traces and replay them with | |
1046 | independent fio invocations. Unfortuantely this also breaks | |
1047 | the strict time ordering between multiple device accesses. | |
1048 | ||
e3cedca7 | 1049 | write_bw_log=str If given, write a bandwidth log of the jobs in this job |
71bfa161 | 1050 | file. Can be used to store data of the bandwidth of the |
e0da9bc2 JA |
1051 | jobs in their lifetime. The included fio_generate_plots |
1052 | script uses gnuplot to turn these text files into nice | |
e3cedca7 JA |
1053 | graphs. See write_log_log for behaviour of given |
1054 | filename. For this option, the postfix is _bw.log. | |
71bfa161 | 1055 | |
e3cedca7 | 1056 | write_lat_log=str Same as write_bw_log, except that this option stores io |
02af0988 JA |
1057 | submission, completion, and total latencies instead. If no |
1058 | filename is given with this option, the default filename of | |
1059 | "jobname_type.log" is used. Even if the filename is given, | |
1060 | fio will still append the type of log. So if one specifies | |
e3cedca7 JA |
1061 | |
1062 | write_lat_log=foo | |
1063 | ||
02af0988 JA |
1064 | The actual log names will be foo_slat.log, foo_slat.log, |
1065 | and foo_lat.log. This helps fio_generate_plot fine the logs | |
1066 | automatically. | |
71bfa161 | 1067 | |
f7fa2653 | 1068 | lockmem=int Pin down the specified amount of memory with mlock(2). Can |
71bfa161 JA |
1069 | potentially be used instead of removing memory or booting |
1070 | with less memory to simulate a smaller amount of memory. | |
1071 | ||
1072 | exec_prerun=str Before running this job, issue the command specified | |
1073 | through system(3). | |
1074 | ||
1075 | exec_postrun=str After the job completes, issue the command specified | |
1076 | though system(3). | |
1077 | ||
1078 | ioscheduler=str Attempt to switch the device hosting the file to the specified | |
1079 | io scheduler before running. | |
1080 | ||
1081 | cpuload=int If the job is a CPU cycle eater, attempt to use the specified | |
1082 | percentage of CPU cycles. | |
1083 | ||
1084 | cpuchunks=int If the job is a CPU cycle eater, split the load into | |
26eca2db | 1085 | cycles of the given time. In microseconds. |
71bfa161 | 1086 | |
0a839f30 JA |
1087 | disk_util=bool Generate disk utilization statistics, if the platform |
1088 | supports it. Defaults to on. | |
1089 | ||
02af0988 | 1090 | disable_lat=bool Disable measurements of total latency numbers. Useful |
9520ebb9 JA |
1091 | only for cutting back the number of calls to gettimeofday, |
1092 | as that does impact performance at really high IOPS rates. | |
1093 | Note that to really get rid of a large amount of these | |
1094 | calls, this option must be used with disable_slat and | |
1095 | disable_bw as well. | |
1096 | ||
02af0988 JA |
1097 | disable_clat=bool Disable measurements of completion latency numbers. See |
1098 | disable_lat. | |
1099 | ||
9520ebb9 | 1100 | disable_slat=bool Disable measurements of submission latency numbers. See |
02af0988 | 1101 | disable_slat. |
9520ebb9 JA |
1102 | |
1103 | disable_bw=bool Disable measurements of throughput/bandwidth numbers. See | |
02af0988 | 1104 | disable_lat. |
9520ebb9 | 1105 | |
993bf48b JA |
1106 | gtod_reduce=bool Enable all of the gettimeofday() reducing options |
1107 | (disable_clat, disable_slat, disable_bw) plus reduce | |
1108 | precision of the timeout somewhat to really shrink | |
1109 | the gettimeofday() call count. With this option enabled, | |
1110 | we only do about 0.4% of the gtod() calls we would have | |
1111 | done if all time keeping was enabled. | |
1112 | ||
be4ecfdf JA |
1113 | gtod_cpu=int Sometimes it's cheaper to dedicate a single thread of |
1114 | execution to just getting the current time. Fio (and | |
1115 | databases, for instance) are very intensive on gettimeofday() | |
1116 | calls. With this option, you can set one CPU aside for | |
1117 | doing nothing but logging current time to a shared memory | |
1118 | location. Then the other threads/processes that run IO | |
1119 | workloads need only copy that segment, instead of entering | |
1120 | the kernel with a gettimeofday() call. The CPU set aside | |
1121 | for doing these time calls will be excluded from other | |
1122 | uses. Fio will manually clear it from the CPU mask of other | |
1123 | jobs. | |
a696fa2a | 1124 | |
f2bba182 RR |
1125 | continue_on_error=bool Normally fio will exit the job on the first observed |
1126 | failure. If this option is set, fio will continue the job when | |
1127 | there is a 'non-fatal error' (EIO or EILSEQ) until the runtime | |
1128 | is exceeded or the I/O size specified is completed. If this | |
1129 | option is used, there are two more stats that are appended, | |
1130 | the total error count and the first error. The error field | |
1131 | given in the stats is the first error that was hit during the | |
1132 | run. | |
be4ecfdf | 1133 | |
6adb38a1 JA |
1134 | cgroup=str Add job to this control group. If it doesn't exist, it will |
1135 | be created. The system must have a mounted cgroup blkio | |
1136 | mount point for this to work. If your system doesn't have it | |
1137 | mounted, you can do so with: | |
a696fa2a JA |
1138 | |
1139 | # mount -t cgroup -o blkio none /cgroup | |
1140 | ||
a696fa2a JA |
1141 | cgroup_weight=int Set the weight of the cgroup to this value. See |
1142 | the documentation that comes with the kernel, allowed values | |
1143 | are in the range of 100..1000. | |
71bfa161 | 1144 | |
7de87099 VG |
1145 | cgroup_nodelete=bool Normally fio will delete the cgroups it has created after |
1146 | the job completion. To override this behavior and to leave | |
1147 | cgroups around after the job completion, set cgroup_nodelete=1. | |
1148 | This can be useful if one wants to inspect various cgroup | |
1149 | files after job completion. Default: false | |
1150 | ||
e0b0d892 JA |
1151 | uid=int Instead of running as the invoking user, set the user ID to |
1152 | this value before the thread/process does any work. | |
1153 | ||
1154 | gid=int Set group ID, see uid. | |
1155 | ||
71bfa161 JA |
1156 | 6.0 Interpreting the output |
1157 | --------------------------- | |
1158 | ||
1159 | fio spits out a lot of output. While running, fio will display the | |
1160 | status of the jobs created. An example of that would be: | |
1161 | ||
73c8b082 | 1162 | Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s] |
71bfa161 JA |
1163 | |
1164 | The characters inside the square brackets denote the current status of | |
1165 | each thread. The possible values (in typical life cycle order) are: | |
1166 | ||
1167 | Idle Run | |
1168 | ---- --- | |
1169 | P Thread setup, but not started. | |
1170 | C Thread created. | |
1171 | I Thread initialized, waiting. | |
b0f65863 | 1172 | p Thread running pre-reading file(s). |
71bfa161 JA |
1173 | R Running, doing sequential reads. |
1174 | r Running, doing random reads. | |
1175 | W Running, doing sequential writes. | |
1176 | w Running, doing random writes. | |
1177 | M Running, doing mixed sequential reads/writes. | |
1178 | m Running, doing mixed random reads/writes. | |
1179 | F Running, currently waiting for fsync() | |
fc6bd43c | 1180 | V Running, doing verification of written data. |
71bfa161 JA |
1181 | E Thread exited, not reaped by main thread yet. |
1182 | _ Thread reaped. | |
1183 | ||
1184 | The other values are fairly self explanatory - number of threads | |
c9f60304 JA |
1185 | currently running and doing io, rate of io since last check (read speed |
1186 | listed first, then write speed), and the estimated completion percentage | |
1187 | and time for the running group. It's impossible to estimate runtime of | |
1188 | the following groups (if any). | |
71bfa161 JA |
1189 | |
1190 | When fio is done (or interrupted by ctrl-c), it will show the data for | |
1191 | each thread, group of threads, and disks in that order. For each data | |
1192 | direction, the output looks like: | |
1193 | ||
1194 | Client1 (g=0): err= 0: | |
b22989b9 | 1195 | write: io= 32MB, bw= 666KB/s, runt= 50320msec |
6104ddb6 JA |
1196 | slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92 |
1197 | clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82 | |
b22989b9 | 1198 | bw (KB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68 |
e7823a94 | 1199 | cpu : usr=1.49%, sys=0.25%, ctx=7969, majf=0, minf=17 |
71619dc2 | 1200 | IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0% |
838bc709 JA |
1201 | submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% |
1202 | complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% | |
30061b97 | 1203 | issued r/w: total=0/32768, short=0/0 |
8abdce66 JA |
1204 | lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%, |
1205 | lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0% | |
71bfa161 JA |
1206 | |
1207 | The client number is printed, along with the group id and error of that | |
1208 | thread. Below is the io statistics, here for writes. In the order listed, | |
1209 | they denote: | |
1210 | ||
1211 | io= Number of megabytes io performed | |
1212 | bw= Average bandwidth rate | |
1213 | runt= The runtime of that thread | |
72fbda2a | 1214 | slat= Submission latency (avg being the average, stdev being the |
71bfa161 JA |
1215 | standard deviation). This is the time it took to submit |
1216 | the io. For sync io, the slat is really the completion | |
8a35c71e | 1217 | latency, since queue/complete is one operation there. This |
bf9a3edb | 1218 | value can be in milliseconds or microseconds, fio will choose |
8a35c71e | 1219 | the most appropriate base and print that. In the example |
bf9a3edb | 1220 | above, milliseconds is the best scale. |
71bfa161 JA |
1221 | clat= Completion latency. Same names as slat, this denotes the |
1222 | time from submission to completion of the io pieces. For | |
1223 | sync io, clat will usually be equal (or very close) to 0, | |
1224 | as the time from submit to complete is basically just | |
1225 | CPU time (io has already been done, see slat explanation). | |
1226 | bw= Bandwidth. Same names as the xlat stats, but also includes | |
1227 | an approximate percentage of total aggregate bandwidth | |
1228 | this thread received in this group. This last value is | |
1229 | only really useful if the threads in this group are on the | |
1230 | same disk, since they are then competing for disk access. | |
1231 | cpu= CPU usage. User and system time, along with the number | |
e7823a94 JA |
1232 | of context switches this thread went through, usage of |
1233 | system and user time, and finally the number of major | |
1234 | and minor page faults. | |
71619dc2 JA |
1235 | IO depths= The distribution of io depths over the job life time. The |
1236 | numbers are divided into powers of 2, so for example the | |
1237 | 16= entries includes depths up to that value but higher | |
1238 | than the previous entry. In other words, it covers the | |
1239 | range from 16 to 31. | |
838bc709 JA |
1240 | IO submit= How many pieces of IO were submitting in a single submit |
1241 | call. Each entry denotes that amount and below, until | |
1242 | the previous entry - eg, 8=100% mean that we submitted | |
1243 | anywhere in between 5-8 ios per submit call. | |
1244 | IO complete= Like the above submit number, but for completions instead. | |
30061b97 JA |
1245 | IO issued= The number of read/write requests issued, and how many |
1246 | of them were short. | |
ec118304 JA |
1247 | IO latencies= The distribution of IO completion latencies. This is the |
1248 | time from when IO leaves fio and when it gets completed. | |
1249 | The numbers follow the same pattern as the IO depths, | |
1250 | meaning that 2=1.6% means that 1.6% of the IO completed | |
8abdce66 JA |
1251 | within 2 msecs, 20=12.8% means that 12.8% of the IO |
1252 | took more than 10 msecs, but less than (or equal to) 20 msecs. | |
71bfa161 JA |
1253 | |
1254 | After each client has been listed, the group statistics are printed. They | |
1255 | will look like this: | |
1256 | ||
1257 | Run status group 0 (all jobs): | |
b22989b9 JA |
1258 | READ: io=64MB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec |
1259 | WRITE: io=64MB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec | |
71bfa161 JA |
1260 | |
1261 | For each data direction, it prints: | |
1262 | ||
1263 | io= Number of megabytes io performed. | |
1264 | aggrb= Aggregate bandwidth of threads in this group. | |
1265 | minb= The minimum average bandwidth a thread saw. | |
1266 | maxb= The maximum average bandwidth a thread saw. | |
1267 | mint= The smallest runtime of the threads in that group. | |
1268 | maxt= The longest runtime of the threads in that group. | |
1269 | ||
1270 | And finally, the disk statistics are printed. They will look like this: | |
1271 | ||
1272 | Disk stats (read/write): | |
1273 | sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00% | |
1274 | ||
1275 | Each value is printed for both reads and writes, with reads first. The | |
1276 | numbers denote: | |
1277 | ||
1278 | ios= Number of ios performed by all groups. | |
1279 | merge= Number of merges io the io scheduler. | |
1280 | ticks= Number of ticks we kept the disk busy. | |
1281 | io_queue= Total time spent in the disk queue. | |
1282 | util= The disk utilization. A value of 100% means we kept the disk | |
1283 | busy constantly, 50% would be a disk idling half of the time. | |
1284 | ||
1285 | ||
1286 | 7.0 Terse output | |
1287 | ---------------- | |
1288 | ||
1289 | For scripted usage where you typically want to generate tables or graphs | |
6af019c9 | 1290 | of the results, fio can output the results in a semicolon separated format. |
71bfa161 JA |
1291 | The format is one long line of values, such as: |
1292 | ||
562c2d2f DN |
1293 | 2;card0;0;0;7139336;121836;60004;1;10109;27.932460;116.933948;220;126861;3495.446807;1085.368601;226;126864;3523.635629;1089.012448;24063;99944;50.275485%;59818.274627;5540.657370;7155060;122104;60004;1;8338;29.086342;117.839068;388;128077;5032.488518;1234.785715;391;128085;5061.839412;1236.909129;23436;100928;50.287926%;59964.832030;5644.844189;14.595833%;19.394167%;123706;0;7313;0.1%;0.1%;0.1%;0.1%;0.1%;0.1%;100.0%;0.00%;0.00%;0.00%;0.00%;0.00%;0.00%;0.01%;0.02%;0.05%;0.16%;6.04%;40.40%;52.68%;0.64%;0.01%;0.00%;0.01%;0.00%;0.00%;0.00%;0.00%;0.00% |
1294 | A description of this job goes here. | |
1295 | ||
1296 | The job description (if provided) follows on a second line. | |
71bfa161 | 1297 | |
525c2bfa JA |
1298 | To enable terse output, use the --minimal command line option. The first |
1299 | value is the version of the terse output format. If the output has to | |
1300 | be changed for some reason, this number will be incremented by 1 to | |
1301 | signify that change. | |
6820cb3b | 1302 | |
71bfa161 JA |
1303 | Split up, the format is as follows: |
1304 | ||
525c2bfa | 1305 | version, jobname, groupid, error |
71bfa161 | 1306 | READ status: |
b22989b9 | 1307 | KB IO, bandwidth (KB/sec), runtime (msec) |
71bfa161 JA |
1308 | Submission latency: min, max, mean, deviation |
1309 | Completion latency: min, max, mean, deviation | |
525c2bfa | 1310 | Total latency: min, max, mean, deviation |
6c219763 | 1311 | Bw: min, max, aggregate percentage of total, mean, deviation |
71bfa161 | 1312 | WRITE status: |
b22989b9 | 1313 | KB IO, bandwidth (KB/sec), runtime (msec) |
71bfa161 JA |
1314 | Submission latency: min, max, mean, deviation |
1315 | Completion latency: min, max, mean, deviation | |
525c2bfa | 1316 | Total latency: min, max, mean, deviation |
6c219763 | 1317 | Bw: min, max, aggregate percentage of total, mean, deviation |
046ee302 | 1318 | CPU usage: user, system, context switches, major faults, minor faults |
2270890c | 1319 | IO depths: <=1, 2, 4, 8, 16, 32, >=64 |
562c2d2f DN |
1320 | IO latencies microseconds: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000 |
1321 | IO latencies milliseconds: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, 2000, >=2000 | |
1322 | Additional Info (dependant on continue_on_error, default off): total # errors, first error code | |
1323 | ||
f42195a3 | 1324 | Additional Info (dependant on description being set): Text description |