4 fio is a tool that will spawn a number of thread doing a particular
5 type of io action as specified by the user. fio takes a number of
6 global parameters, each inherited by the thread unless otherwise
7 parameters given to them overriding that setting is given.
13 fio resides in a git repo, the canonical place is:
15 git://brick.kernel.dk/data/git/fio.git
17 Snapshots are frequently generated as well and they include the git
18 meta data as well. You can download them here:
20 http://brick.kernel.dk/snaps/
22 Pascal Bleser <guru@unixtech.be> has fio RPMs in his repository, you
25 http://linux01.gwdg.de/~pbleser/rpm-navigation.php?cat=System/fio
33 -b block size in KiB for each io
34 -t <sec> Runtime in seconds
35 -r For random io, sequence must be repeatable
36 -R <on> If one thread fails to meet rate, quit all
37 -o <on> Use direct IO is 1, buffered if 0
38 -l Generate per-job latency logs
39 -w Generate per-job bandwidth logs
40 -f <file> Read <file> for job descriptions
42 -v Print version information and exit
44 The <jobs> format is as follows:
46 name=x Use 'x' as the identifier for this job.
47 directory=x Use 'x' as the top level directory for storing files
48 rw=x 'x' may be: read, randread, write, randwrite,
49 rw (read-write mix), randrw (read-write random mix)
50 rwmixcycle=x Base cycle for switching between read and write
52 rwmixread=x 'x' percentage of rw mix ios will be reads. If
53 rwmixwrite is also given, the last of the two will
54 be used if they don't add up to 100%.
55 rwmixwrite=x 'x' percentage of rw mix ios will be writes. See
57 size=x Set file size to x bytes (x string can include k/m/g)
58 ioengine=x 'x' may be: aio/libaio/linuxaio for Linux aio,
59 posixaio for POSIX aio, sync for regular read/write io,
60 mmap for mmap'ed io, splice for using splice/vmsplice,
61 or sgio for direct SG_IO io. The latter only works on
62 Linux on SCSI (or SCSI-like devices, such as
63 usb-storage or sata/libata driven) devices.
64 iodepth=x For async io, allow 'x' ios in flight
65 overwrite=x If 'x', layout a write file first.
66 prio=x Run io at prio X, 0-7 is the kernel allowed range
67 prioclass=x Run io at prio class X
68 bs=x Use 'x' for thread blocksize. May include k/m postfix.
69 bsrange=x-y Mix thread block sizes randomly between x and y. May
70 also include k/m postfix.
71 direct=x 1 for direct IO, 0 for buffered IO
72 thinktime=x "Think" x usec after each io
73 rate=x Throttle rate to x KiB/sec
74 ratemin=x Quit if rate of x KiB/sec can't be met
75 ratecycle=x ratemin averaged over x msecs
76 cpumask=x Only allow job to run on CPUs defined by mask.
77 fsync=x If writing, fsync after every x blocks have been written
78 startdelay=x Start this thread x seconds after startup
79 timeout=x Terminate x seconds after startup
80 offset=x Start io at offset x (x string can include k/m/g)
81 invalidate=x Invalidate page cache for file prior to doing io
82 sync=x Use sync writes if x and writing
83 mem=x If x == malloc, use malloc for buffers. If x == shm,
84 use shm for buffers. If x == mmap, use anon mmap.
85 exitall When one thread quits, terminate the others
86 bwavgtime=x Average bandwidth stats over an x msec window.
87 create_serialize=x If 'x', serialize file creation.
88 create_fsync=x If 'x', run fsync() after file creation.
89 end_fsync=x If 'x', run fsync() after end-of-job.
90 loops=x Run the job 'x' number of times.
91 verify=x If 'x' == md5, use md5 for verifies. If 'x' == crc32,
92 use crc32 for verifies. md5 is 'safer', but crc32 is
93 a lot faster. Only makes sense for writing to a file.
94 stonewall Wait for preceeding jobs to end before running.
95 numjobs=x Create 'x' similar entries for this job
96 thread Use pthreads instead of forked jobs
98 zoneskip=y Zone options must be paired. If given, the job
99 will skip y bytes for every x read/written. This
100 can be used to gauge hard drive speed over the entire
101 platter, without reading everything. Both x/y can
102 include k/m/g suffix.
103 iolog=x Open and read io pattern from file 'x'. The file must
104 contain one io action per line in the following format:
106 where with rw=0/1 for read/write, and the offset
107 and length entries being in bytes.
108 write_iolog=x Write an iolog to file 'x' in the same format as iolog.
109 The iolog options are exclusive, if both given the
110 read iolog will be performed.
111 lockmem=x Lock down x amount of memory on the machine, to
112 simulate a machine with less memory available. x can
113 include k/m/g suffix.
114 nice=x Run job at given nice value.
115 exec_prerun=x Run 'x' before job io is begun.
116 exec_postrun=x Run 'x' after job io has finished.
117 ioscheduler=x Use ioscheduler 'x' for this job.
119 Examples using a job file
120 -------------------------
122 A sample job file doing the same as above would look like this:
132 And fio would be invoked as:
134 $ fio -o1 -s -f file_with_above
136 The second example would look like this:
151 And fio would be invoked as:
153 $ fio -o0 -s -b4096 -f file_with_above
155 'global' is a reserved keyword. When used as the filename, it sets the
156 default options for the threads following that section. It is possible
157 to have more than one global section in the file, as it only affects
160 Also see the examples/ dir for sample job files.
163 Interpreting the output
164 -----------------------
166 fio spits out a lot of output. While running, fio will display the
167 status of the jobs created. An example of that would be:
169 Threads now running: 2 : [ww] [5.73% done]
171 The characters inside the square brackets denote the current status of
172 each thread. The possible values (in typical life cycle order) are:
176 P Thread setup, but not started.
177 C Thread created and running, but not doing anything yet
178 R Running, doing sequential reads.
179 r Running, doing random reads.
180 W Running, doing sequential writes.
181 w Running, doing random writes.
182 V Running, doing verification of written data.
183 E Thread exited, not reaped by main thread yet.
186 The other values are fairly self explanatory - number of thread currently
187 running and doing io, and the estimated completion percentage.
189 When fio is done (or interrupted by ctrl-c), it will show the data for
190 each thread, group of threads, and disks in that order. For each data
191 direction, the output looks like:
193 Client1 (g=0): err= 0:
194 write: io= 32MiB, bw= 666KiB/s, runt= 50320msec
195 slat (msec): min= 0, max= 136, avg= 0.03, dev= 1.92
196 clat (msec): min= 0, max= 631, avg=48.50, dev=86.82
197 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, dev=681.68
198 cpu : usr=1.49%, sys=0.25%, ctx=7969
200 The client number is printed, along with the group id and error of that
201 thread. Below is the io statistics, here for writes. In the order listed,
204 io= Number of megabytes io performed
205 bw= Average bandwidth rate
206 runt= The runtime of that thread
207 slat= Submission latency (avg being the average, dev being the
208 standard deviation). This is the time it took to submit
209 the io. For sync io, the slat is really the completion
210 latency, since queue/complete is one operation there.
211 clat= Completion latency. Same names as slat, this denotes the
212 time from submission to completion of the io pieces. For
213 sync io, clat will usually be equal (or very close) to 0,
214 as the time from submit to complete is basically just
215 CPU time (io has already been done, see slat explanation).
216 bw= Bandwidth. Same names as the xlat stats, but also includes
217 an approximate percentage of total aggregate bandwidth
218 this thread received in this group. This last value is
219 only really useful if the threads in this group are on the
220 same disk, since they are then competing for disk access.
221 cpu= CPU usage. User and system time, along with the number
222 of context switches this thread went through.
224 After each client has been listed, the group statistics are printed. They
227 Run status group 0 (all jobs):
228 READ: io=64MiB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
229 WRITE: io=64MiB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
231 For each data direction, it prints:
233 io= Number of megabytes io performed.
234 aggrb= Aggregate bandwidth of threads in this group.
235 minb= The minimum average bandwidth a thread saw.
236 maxb= The maximum average bandwidth a thread saw.
237 mint= The minimum runtime of a thread.
238 maxt= The maximum runtime of a thread.
240 And finally, the disk statistics are printed. They will look like this:
242 Disk stats (read/write):
243 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
245 Each value is printed for both reads and writes, with reads first. The
248 ios= Number of ios performed by all groups.
249 merge= Number of merges io the io scheduler.
250 ticks= Number of ticks we kept the disk busy.
251 io_queue= Total time spent in the disk queue.
252 util= The disk utilization. A value of 100% means we kept the disk
253 busy constantly, 50% would be a disk idling half of the time.