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 directory=x Use 'x' as the top level directory for storing files
47 rw=x 'x' may be: read, randread, write, randwrite,
48 rw (read-write mix), randrw (read-write random mix)
49 rwmixcycle=x Base cycle for switching between read and write
51 rwmixread=x 'x' percentage of rw mix ios will be reads. If
52 rwmixwrite is also given, the last of the two will
53 be used if they don't add up to 100%.
54 rwmixwrite=x 'x' percentage of rw mix ios will be writes. See
56 size=x Set file size to x bytes (x string can include k/m/g)
57 ioengine=x 'x' may be: aio/libaio/linuxaio for Linux aio,
58 posixaio for POSIX aio, sync for regular read/write io,
59 mmap for mmap'ed io, splice for using splice/vmsplice,
60 or sgio for direct SG_IO io. The latter only works on
61 Linux on SCSI (or SCSI-like devices, such as
62 usb-storage or sata/libata driven) devices.
63 iodepth=x For async io, allow 'x' ios in flight
64 overwrite=x If 'x', layout a write file first.
65 prio=x Run io at prio X, 0-7 is the kernel allowed range
66 prioclass=x Run io at prio class X
67 bs=x Use 'x' for thread blocksize. May include k/m postfix.
68 bsrange=x-y Mix thread block sizes randomly between x and y. May
69 also include k/m postfix.
70 direct=x 1 for direct IO, 0 for buffered IO
71 thinktime=x "Think" x usec after each io
72 rate=x Throttle rate to x KiB/sec
73 ratemin=x Quit if rate of x KiB/sec can't be met
74 ratecycle=x ratemin averaged over x msecs
75 cpumask=x Only allow job to run on CPUs defined by mask.
76 fsync=x If writing, fsync after every x blocks have been written
77 startdelay=x Start this thread x seconds after startup
78 timeout=x Terminate x seconds after startup
79 offset=x Start io at offset x (x string can include k/m/g)
80 invalidate=x Invalidate page cache for file prior to doing io
81 sync=x Use sync writes if x and writing
82 mem=x If x == malloc, use malloc for buffers. If x == shm,
83 use shm for buffers. If x == mmap, use anon mmap.
84 exitall When one thread quits, terminate the others
85 bwavgtime=x Average bandwidth stats over an x msec window.
86 create_serialize=x If 'x', serialize file creation.
87 create_fsync=x If 'x', run fsync() after file creation.
88 end_fsync=x If 'x', run fsync() after end-of-job.
89 loops=x Run the job 'x' number of times.
90 verify=x If 'x' == md5, use md5 for verifies. If 'x' == crc32,
91 use crc32 for verifies. md5 is 'safer', but crc32 is
92 a lot faster. Only makes sense for writing to a file.
93 stonewall Wait for preceeding jobs to end before running.
94 numjobs=x Create 'x' similar entries for this job
95 thread Use pthreads instead of forked jobs
97 zoneskip=y Zone options must be paired. If given, the job
98 will skip y bytes for every x read/written. This
99 can be used to gauge hard drive speed over the entire
100 platter, without reading everything. Both x/y can
101 include k/m/g suffix.
102 iolog=x Open and read io pattern from file 'x'. The file must
103 contain one io action per line in the following format:
105 where with rw=0/1 for read/write, and the offset
106 and length entries being in bytes.
107 write_iolog=x Write an iolog to file 'x' in the same format as iolog.
108 The iolog options are exclusive, if both given the
109 read iolog will be performed.
110 lockmem=x Lock down x amount of memory on the machine, to
111 simulate a machine with less memory available. x can
112 include k/m/g suffix.
113 nice=x Run job at given nice value.
114 exec_prerun=x Run 'x' before job io is begun.
115 exec_postrun=x Run 'x' after job io has finished.
117 Examples using a job file
118 -------------------------
120 A sample job file doing the same as above would look like this:
130 And fio would be invoked as:
132 $ fio -o1 -s -f file_with_above
134 The second example would look like this:
149 And fio would be invoked as:
151 $ fio -o0 -s -b4096 -f file_with_above
153 'global' is a reserved keyword. When used as the filename, it sets the
154 default options for the threads following that section. It is possible
155 to have more than one global section in the file, as it only affects
158 Also see the examples/ dir for sample job files.
161 Interpreting the output
162 -----------------------
164 fio spits out a lot of output. While running, fio will display the
165 status of the jobs created. An example of that would be:
167 Threads now running: 2 : [ww] [5.73% done]
169 The characters inside the square brackets denote the current status of
170 each thread. The possible values (in typical life cycle order) are:
174 P Thread setup, but not started.
175 C Thread created and running, but not doing anything yet
176 R Running, doing sequential reads.
177 r Running, doing random reads.
178 W Running, doing sequential writes.
179 w Running, doing random writes.
180 V Running, doing verification of written data.
181 E Thread exited, not reaped by main thread yet.
184 The other values are fairly self explanatory - number of thread currently
185 running and doing io, and the estimated completion percentage.
187 When fio is done (or interrupted by ctrl-c), it will show the data for
188 each thread, group of threads, and disks in that order. For each data
189 direction, the output looks like:
191 Client1 (g=0): err= 0:
192 write: io= 32MiB, bw= 666KiB/s, runt= 50320msec
193 slat (msec): min= 0, max= 136, avg= 0.03, dev= 1.92
194 clat (msec): min= 0, max= 631, avg=48.50, dev=86.82
195 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, dev=681.68
196 cpu : usr=1.49%, sys=0.25%, ctx=7969
198 The client number is printed, along with the group id and error of that
199 thread. Below is the io statistics, here for writes. In the order listed,
202 io= Number of megabytes io performed
203 bw= Average bandwidth rate
204 runt= The runtime of that thread
205 slat= Submission latency (avg being the average, dev being the
206 standard deviation). This is the time it took to submit
207 the io. For sync io, the slat is really the completion
208 latency, since queue/complete is one operation there.
209 clat= Completion latency. Same names as slat, this denotes the
210 time from submission to completion of the io pieces. For
211 sync io, clat will usually be equal (or very close) to 0,
212 as the time from submit to complete is basically just
213 CPU time (io has already been done, see slat explanation).
214 bw= Bandwidth. Same names as the xlat stats, but also includes
215 an approximate percentage of total aggregate bandwidth
216 this thread received in this group. This last value is
217 only really useful if the threads in this group are on the
218 same disk, since they are then competing for disk access.
219 cpu= CPU usage. User and system time, along with the number
220 of context switches this thread went through.
222 After each client has been listed, the group statistics are printed. They
225 Run status group 0 (all jobs):
226 READ: io=64MiB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
227 WRITE: io=64MiB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
229 For each data direction, it prints:
231 io= Number of megabytes io performed.
232 aggrb= Aggregate bandwidth of threads in this group.
233 minb= The minimum average bandwidth a thread saw.
234 maxb= The maximum average bandwidth a thread saw.
235 mint= The minimum runtime of a thread.
236 maxt= The maximum runtime of a thread.
238 And finally, the disk statistics are printed. They will look like this:
240 Disk stats (read/write):
241 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
243 Each value is printed for both reads and writes, with reads first. The
246 ios= Number of ios performed by all groups.
247 merge= Number of merges io the io scheduler.
248 ticks= Number of ticks we kept the disk busy.
249 io_queue= Total time spent in the disk queue.
250 util= The disk utilization. A value of 100% means we kept the disk
251 busy constantly, 50% would be a disk idling half of the time.