[PATCH] Fixup iolog config setting
[fio.git] / HOWTO
1Table of contents
41. Overview
52. How fio works
63. Running fio
74. Job file format
85. Detailed list of parameters
96. Normal output
107. Terse output
131.0 Overview and history
15fio was originally written to save me the hassle of writing special test
16case programs when I wanted to test a specific workload, either for
17performance reasons or to find/reproduce a bug. The process of writing
18such a test app can be tiresome, especially if you have to do it often.
19Hence I needed a tool that would be able to simulate a given io workload
20without resorting to writing a tailored test case again and again.
22A test work load is difficult to define, though. There can be any number
23of processes or threads involved, and they can each be using their own
24way of generating io. You could have someone dirtying large amounts of
25memory in an memory mapped file, or maybe several threads issuing
26reads using asynchronous io. fio needed to be flexible enough to
27simulate both of these cases, and many more.
292.0 How fio works
31The first step in getting fio to simulate a desired io workload, is
32writing a job file describing that specific setup. A job file may contain
33any number of threads and/or files - the typical contents of the job file
34is a global section defining shared parameters, and one or more job
35sections describing the jobs involved. When run, fio parses this file
36and sets everything up as described. If we break down a job from top to
37bottom, it contains the following basic parameters:
39 IO type Defines the io pattern issued to the file(s).
40 We may only be reading sequentially from this
41 file(s), or we may be writing randomly. Or even
42 mixing reads and writes, sequentially or randomly.
44 Block size In how large chunks are we issuing io? This may be
45 a single value, or it may describe a range of
46 block sizes.
48 IO size How much data are we going to be reading/writing.
50 IO engine How do we issue io? We could be memory mapping the
51 file, we could be using regular read/write, we
52 could be using splice, async io, or even
53 SG (SCSI generic sg).
55 IO depth If the io engine is async, how large a queueing
56 depth do we want to maintain?
58 IO type Should we be doing buffered io, or direct/raw io?
60 Num files How many files are we spreading the workload over.
62 Num threads How many threads or processes should we spread
63 this workload over.
65The above are the basic parameters defined for a workload, in addition
66there's a multitude of parameters that modify other aspects of how this
67job behaves.
703.0 Running fio
72See the README file for command line parameters, there are only a few
73of them.
75Running fio is normally the easiest part - you just give it the job file
76(or job files) as parameters:
78$ fio job_file
80and it will start doing what the job_file tells it to do. You can give
81more than one job file on the command line, fio will serialize the running
82of those files. Internally that is the same as using the 'stonewall'
83parameter described the the parameter section.
85fio does not need to run as root, except if the files or devices specified
86in the job section requires that. Some other options may also be restricted,
87such as memory locking, io scheduler switching, and descreasing the nice value.
904.0 Job file format
92As previously described, fio accepts one or more job files describing
93what it is supposed to do. The job file format is the classic ini file,
94where the names enclosed in [] brackets define the job name. You are free
95to use any ascii name you want, except 'global' which has special meaning.
96A global section sets defaults for the jobs described in that file. A job
97may override a global section parameter, and a job file may even have
98several global sections if so desired. A job is only affected by a global
99section residing above it. If the first character in a line is a ';', the
100entire line is discarded as a comment.
102So lets look at a really simple job file that define to threads, each
103randomly reading from a 128MiB file.
105; -- start job file --
114; -- end job file --
116As you can see, the job file sections themselves are empty as all the
117described parameters are shared. As no filename= option is given, fio
118makes up a filename for each of the jobs as it sees fit.
120Lets look at an example that have a number of processes writing randomly
121to files.
123; -- start job file --
133; -- end job file --
135Here we have no global section, as we only have one job defined anyway.
136We want to use async io here, with a depth of 4 for each file. We also
137increased the buffer size used to 32KiB and define numjobs to 4 to
138fork 4 identical jobs. The result is 4 processes each randomly writing
139to their own 64MiB file.
141fio ships with a few example job files, you can also look there for
1455.0 Detailed list of parameters
148This section describes in details each parameter associated with a job.
149Some parameters take an option of a given type, such as an integer or
150a string. The following types are used:
152str String. This is a sequence of alpha characters.
153int Integer. A whole number value, may be negative.
154siint SI integer. A whole number value, which may contain a postfix
155 describing the base of the number. Accepted postfixes are k/m/g,
156 meaning kilo, mega, and giga. So if you want to specifiy 4096,
157 you could either write out '4096' or just give 4k. The postfixes
158 signify base 2 values, so 1024 is 1k and 1024k is 1m and so on.
159bool Boolean. Usually parsed as an integer, however only defined for
160 true and false (1 and 0).
161irange Integer range with postfix. Allows value range to be given, such
162 as 1024-4096. Also see siint.
164With the above in mind, here follows the complete list of fio job
167name=str ASCII name of the job. This may be used to override the
168 name printed by fio for this job. Otherwise the job
169 name is used.
171directory=str Prefix filenames with this directory. Used to places files
172 in a different location than "./".
174filename=str Fio normally makes up a filename based on the job name,
175 thread number, and file number. If you want to share
176 files between threads in a job or several jobs, specify
177 a filename for each of them to override the default.
179rw=str Type of io pattern. Accepted values are:
181 read Sequential reads
182 write Sequential writes
183 randwrite Random writes
184 randread Random reads
185 rw Sequential mixed reads and writes
186 randrw Random mixed reads and writes
188 For the mixed io types, the default is to split them 50/50.
189 For certain types of io the result may still be skewed a bit,
190 since the speed may be different.
192size=siint The total size of file io for this job. This may describe
193 the size of the single file the job uses, or it may be
194 divided between the number of files in the job. If the
195 file already exists, the file size will be adjusted to this
196 size if larger than the current file size. If this parameter
197 is not given and the file exists, the file size will be used.
199bs=siint The block size used for the io units. Defaults to 4k.
201bsrange=irange Instead of giving a single block size, specify a range
202 and fio will mix the issued io block sizes. The issued
203 io unit will always be a multiple of the minimum value
204 given.
206nrfiles=int Number of files to use for this job. Defaults to 1.
208ioengine=str Defines how the job issues io to the file. The following
209 types are defined:
211 sync Basic read(2) or write(2) io. lseek(2) is
212 used to position the io location.
214 libaio Linux native asynchronous io.
216 posixaio glibc posix asynchronous io.
218 mmap File is memory mapped and data copied
219 to/from using memcpy(3).
221 splice splice(2) is used to transfer the data and
222 vmsplice(2) to transfer data from user
223 space to the kernel.
225 sg SCSI generic sg v3 io. May either be
226 syncrhonous using the SG_IO ioctl, or if
227 the target is an sg character device
228 we use read(2) and write(2) for asynchronous
229 io.
231iodepth=int This defines how many io units to keep in flight against
232 the file. The default is 1 for each file defined in this
233 job, can be overridden with a larger value for higher
234 concurrency.
236direct=bool If value is true, use non-buffered io. This is usually
237 O_DIRECT. Defaults to true.
239offset=siint Start io at the given offset in the file. The data before
240 the given offset will not be touched. This effectively
241 caps the file size at real_size - offset.
243fsync=int If writing to a file, issue a sync of the dirty data
244 for every number of blocks given. For example, if you give
245 32 as a parameter, fio will sync the file for every 32
246 writes issued. If fio is using non-buffered io, we may
247 not sync the file. The exception is the sg io engine, which
248 syncronizes the disk cache anyway.
250overwrite=bool If writing to a file, setup the file first and do overwrites.
252end_fsync=bool If true, fsync file contents when the job exits.
254rwmixcycle=int Value in miliseconds describing how often to switch between
255 reads and writes for a mixed workload. The default is
256 500 msecs.
258rwmixread=int How large a percentage of the mix should be reads.
260rwmixwrite=int How large a percentage of the mix should be writes. If both
261 rwmixread and rwmixwrite is given and the values do not add
262 up to 100%, the latter of the two will be used to override
263 the first.
265nice=int Run the job with the given nice value. See man nice(2).
267prio=int Set the io priority value of this job. Linux limits us to
268 a positive value between 0 and 7, with 0 being the highest.
269 See man ionice(1).
271prioclass=int Set the io priority class. See man ionice(1).
273thinktime=int Stall the job x microseconds after an io has completed before
274 issuing the next. May be used to simulate processing being
275 done by an application.
277rate=int Cap the bandwidth used by this job to this number of KiB/sec.
279ratemin=int Tell fio to do whatever it can to maintain at least this
280 bandwidth.
282ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
283 of miliseconds.
285cpumask=int Set the CPU affinity of this job. The parameter given is a
286 bitmask of allowed CPU's the job may run on. See man
287 sched_setaffinity(2).
289startdelay=int Start this job the specified number of seconds after fio
290 has started. Only useful if the job file contains several
291 jobs, and you want to delay starting some jobs to a certain
292 time.
294timeout=int Tell fio to terminate processing after the specified number
295 of seconds. It can be quite hard to determine for how long
296 a specified job will run, so this parameter is handy to
297 cap the total runtime to a given time.
299invalidate=bool Invalidate the buffer/page cache parts for this file prior
300 to starting io. Defaults to true.
302sync=bool Use sync io for buffered writes. For the majority of the
303 io engines, this means using O_SYNC.
305mem=str Fio can use various types of memory as the io unit buffer.
306 The allowed values are:
308 malloc Use memory from malloc(3) as the buffers.
310 shm Use shared memory as the buffers. Allocated
311 through shmget(2).
313 mmap Use anonymous memory maps as the buffers.
314 Allocated through mmap(2).
316 The area allocated is a function of the maximum allowed
317 bs size for the job, multiplied by the io depth given.
319exitall When one job finishes, terminate the rest. The default is
320 to wait for each job to finish, sometimes that is not the
321 desired action.
323bwavgtime=int Average the calculated bandwidth over the given time. Value
324 is specified in miliseconds.
326create_serialize=bool If true, serialize the file creating for the jobs.
327 This may be handy to avoid interleaving of data
328 files, which may greatly depend on the filesystem
329 used and even the number of processors in the system.
331create_fsync=bool fsync the data file after creation. This is the
332 default.
334unlink Unlink the job files when done. fio defaults to doing this,
335 if it created the file itself.
337loops=int Run the specified number of iterations of this job. Used
338 to repeat the same workload a given number of times. Defaults
339 to 1.
341verify=str If writing to a file, fio can verify the file contents
342 after each iteration of the job. The allowed values are:
344 md5 Use an md5 sum of the data area and store
345 it in the header of each block.
347 crc32 Use a crc32 sum of the data area and store
348 it in the header of each block.
350 This option can be used for repeated burnin tests of a
351 system to make sure that the written data is also
352 correctly read back.
354stonewall Wait for preceeding jobs in the job file to exit, before
355 starting this one. Can be used to insert serialization
356 points in the job file.
358numjobs=int Create the specified number of clones of this job. May be
359 used to setup a larger number of threads/processes doing
360 the same thing.
362thread fio defaults to forking jobs, however if this option is
363 given, fio will use pthread_create(3) to create threads
364 instead.
366zonesize=siint Divide a file into zones of the specified size. See zoneskip.
368zoneskip=siint Skip the specified number of bytes when zonesize data has
369 been read. The two zone options can be used to only do
370 io on zones of a file.
372write_iolog=str Write the issued io patterns to the specified file. See iolog.
374iolog=str Open an iolog with the specified file name and replay the
375 io patterns it contains. This can be used to store a
376 workload and replay it sometime later.
378write_bw_log If given, write a bandwidth log of the jobs in this job
379 file. Can be used to store data of the bandwidth of the
380 jobs in their lifetime. The included fio_generate_plots
381 script uses gnuplot to turn these text files into nice
382 graphs.
384write_lat_log Same as write_bw_log, except that this option stores io
385 completion latencies instead.
387lockmem=siint Pin down the specified amount of memory with mlock(2). Can
388 potentially be used instead of removing memory or booting
389 with less memory to simulate a smaller amount of memory.
391exec_prerun=str Before running this job, issue the command specified
392 through system(3).
394exec_postrun=str After the job completes, issue the command specified
395 though system(3).
397ioscheduler=str Attempt to switch the device hosting the file to the specified
398 io scheduler before running.
400cpuload=int If the job is a CPU cycle eater, attempt to use the specified
401 percentage of CPU cycles.
403cpuchunks=int If the job is a CPU cycle eater, split the load into
404 cycles of the given time. In miliseconds.
4076.0 Interpreting the output
410fio spits out a lot of output. While running, fio will display the
411status of the jobs created. An example of that would be:
413Threads running: 1: [_r] [24.79% done] [eta 00h:01m:31s]
415The characters inside the square brackets denote the current status of
416each thread. The possible values (in typical life cycle order) are:
418Idle Run
419---- ---
420P Thread setup, but not started.
421C Thread created.
422I Thread initialized, waiting.
423 R Running, doing sequential reads.
424 r Running, doing random reads.
425 W Running, doing sequential writes.
426 w Running, doing random writes.
427 M Running, doing mixed sequential reads/writes.
428 m Running, doing mixed random reads/writes.
429 F Running, currently waiting for fsync()
430V Running, doing verification of written data.
431E Thread exited, not reaped by main thread yet.
432_ Thread reaped.
434The other values are fairly self explanatory - number of threads
435currently running and doing io, and the estimated completion percentage
436and time for the running group. It's impossible to estimate runtime
437of the following groups (if any).
439When fio is done (or interrupted by ctrl-c), it will show the data for
440each thread, group of threads, and disks in that order. For each data
441direction, the output looks like:
443Client1 (g=0): err= 0:
444 write: io= 32MiB, bw= 666KiB/s, runt= 50320msec
445 slat (msec): min= 0, max= 136, avg= 0.03, dev= 1.92
446 clat (msec): min= 0, max= 631, avg=48.50, dev=86.82
447 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, dev=681.68
448 cpu : usr=1.49%, sys=0.25%, ctx=7969
450The client number is printed, along with the group id and error of that
451thread. Below is the io statistics, here for writes. In the order listed,
452they denote:
454io= Number of megabytes io performed
455bw= Average bandwidth rate
456runt= The runtime of that thread
457 slat= Submission latency (avg being the average, dev being the
458 standard deviation). This is the time it took to submit
459 the io. For sync io, the slat is really the completion
460 latency, since queue/complete is one operation there.
461 clat= Completion latency. Same names as slat, this denotes the
462 time from submission to completion of the io pieces. For
463 sync io, clat will usually be equal (or very close) to 0,
464 as the time from submit to complete is basically just
465 CPU time (io has already been done, see slat explanation).
466 bw= Bandwidth. Same names as the xlat stats, but also includes
467 an approximate percentage of total aggregate bandwidth
468 this thread received in this group. This last value is
469 only really useful if the threads in this group are on the
470 same disk, since they are then competing for disk access.
471cpu= CPU usage. User and system time, along with the number
472 of context switches this thread went through.
474After each client has been listed, the group statistics are printed. They
475will look like this:
477Run status group 0 (all jobs):
478 READ: io=64MiB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
479 WRITE: io=64MiB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
481For each data direction, it prints:
483io= Number of megabytes io performed.
484aggrb= Aggregate bandwidth of threads in this group.
485minb= The minimum average bandwidth a thread saw.
486maxb= The maximum average bandwidth a thread saw.
487mint= The smallest runtime of the threads in that group.
488maxt= The longest runtime of the threads in that group.
490And finally, the disk statistics are printed. They will look like this:
492Disk stats (read/write):
493 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
495Each value is printed for both reads and writes, with reads first. The
496numbers denote:
498ios= Number of ios performed by all groups.
499merge= Number of merges io the io scheduler.
500ticks= Number of ticks we kept the disk busy.
501io_queue= Total time spent in the disk queue.
502util= The disk utilization. A value of 100% means we kept the disk
503 busy constantly, 50% would be a disk idling half of the time.
5067.0 Terse output
509For scripted usage where you typically want to generate tables or graphs
510of the results, fio can output the results in a comma seperated format.
511The format is one long line of values, such as:
515Split up, the format is as follows:
517 jobname, groupid, error
518 READ status:
519 KiB IO, bandwidth (KiB/sec), runtime (msec)
520 Submission latency: min, max, mean, deviation
521 Completion latency: min, max, mean, deviation
522 Bw: min, max, aggreate percentage of total, mean, deviation
523 WRITE status:
524 KiB IO, bandwidth (KiB/sec), runtime (msec)
525 Submission latency: min, max, mean, deviation
526 Completion latency: min, max, mean, deviation
527 Bw: min, max, aggreate percentage of total, mean, deviation
528 CPU usage: user, system, context switches