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
31 Just type 'make' and 'make install'. If on FreeBSD, for now you have to
32 specify the FreeBSD Makefile with -f, eg:
34 $ make -f Makefile.Freebsd && make -f Makefile.FreeBSD install
36 This might change in the future if I opt for an autoconf type setup.
44 -b block size in KiB for each io
45 -t <sec> Runtime in seconds
46 -r For random io, sequence must be repeatable
47 -R <on> If one thread fails to meet rate, quit all
48 -o <on> Use direct IO is 1, buffered if 0
49 -l Generate per-job latency logs
50 -w Generate per-job bandwidth logs
51 -f <file> Read <file> for job descriptions
53 -v Print version information and exit
55 The <jobs> format is as follows:
57 name=x Use 'x' as the identifier for this job.
58 directory=x Use 'x' as the top level directory for storing files
59 rw=x 'x' may be: read, randread, write, randwrite,
60 rw (read-write mix), randrw (read-write random mix)
61 rwmixcycle=x Base cycle for switching between read and write
63 rwmixread=x 'x' percentage of rw mix ios will be reads. If
64 rwmixwrite is also given, the last of the two will
65 be used if they don't add up to 100%.
66 rwmixwrite=x 'x' percentage of rw mix ios will be writes. See
68 size=x Set file size to x bytes (x string can include k/m/g)
69 ioengine=x 'x' may be: aio/libaio/linuxaio for Linux aio,
70 posixaio for POSIX aio, sync for regular read/write io,
71 mmap for mmap'ed io, splice for using splice/vmsplice,
72 or sgio for direct SG_IO io. The latter only works on
73 Linux on SCSI (or SCSI-like devices, such as
74 usb-storage or sata/libata driven) devices.
75 iodepth=x For async io, allow 'x' ios in flight
76 overwrite=x If 'x', layout a write file first.
77 prio=x Run io at prio X, 0-7 is the kernel allowed range
78 prioclass=x Run io at prio class X
79 bs=x Use 'x' for thread blocksize. May include k/m postfix.
80 bsrange=x-y Mix thread block sizes randomly between x and y. May
81 also include k/m postfix.
82 direct=x 1 for direct IO, 0 for buffered IO
83 thinktime=x "Think" x usec after each io
84 rate=x Throttle rate to x KiB/sec
85 ratemin=x Quit if rate of x KiB/sec can't be met
86 ratecycle=x ratemin averaged over x msecs
87 cpumask=x Only allow job to run on CPUs defined by mask.
88 fsync=x If writing, fsync after every x blocks have been written
89 startdelay=x Start this thread x seconds after startup
90 timeout=x Terminate x seconds after startup
91 offset=x Start io at offset x (x string can include k/m/g)
92 invalidate=x Invalidate page cache for file prior to doing io
93 sync=x Use sync writes if x and writing
94 mem=x If x == malloc, use malloc for buffers. If x == shm,
95 use shm for buffers. If x == mmap, use anon mmap.
96 exitall When one thread quits, terminate the others
97 bwavgtime=x Average bandwidth stats over an x msec window.
98 create_serialize=x If 'x', serialize file creation.
99 create_fsync=x If 'x', run fsync() after file creation.
100 end_fsync=x If 'x', run fsync() after end-of-job.
101 loops=x Run the job 'x' number of times.
102 verify=x If 'x' == md5, use md5 for verifies. If 'x' == crc32,
103 use crc32 for verifies. md5 is 'safer', but crc32 is
104 a lot faster. Only makes sense for writing to a file.
105 stonewall Wait for preceeding jobs to end before running.
106 numjobs=x Create 'x' similar entries for this job
107 thread Use pthreads instead of forked jobs
109 zoneskip=y Zone options must be paired. If given, the job
110 will skip y bytes for every x read/written. This
111 can be used to gauge hard drive speed over the entire
112 platter, without reading everything. Both x/y can
113 include k/m/g suffix.
114 iolog=x Open and read io pattern from file 'x'. The file must
115 contain one io action per line in the following format:
117 where with rw=0/1 for read/write, and the offset
118 and length entries being in bytes.
119 write_iolog=x Write an iolog to file 'x' in the same format as iolog.
120 The iolog options are exclusive, if both given the
121 read iolog will be performed.
122 lockmem=x Lock down x amount of memory on the machine, to
123 simulate a machine with less memory available. x can
124 include k/m/g suffix.
125 nice=x Run job at given nice value.
126 exec_prerun=x Run 'x' before job io is begun.
127 exec_postrun=x Run 'x' after job io has finished.
128 ioscheduler=x Use ioscheduler 'x' for this job.
130 Examples using a job file
131 -------------------------
133 A sample job file doing the same as above would look like this:
143 And fio would be invoked as:
145 $ fio -o1 -s -f file_with_above
147 The second example would look like this:
162 And fio would be invoked as:
164 $ fio -o0 -s -b4096 -f file_with_above
166 'global' is a reserved keyword. When used as the filename, it sets the
167 default options for the threads following that section. It is possible
168 to have more than one global section in the file, as it only affects
171 Also see the examples/ dir for sample job files.
174 Interpreting the output
175 -----------------------
177 fio spits out a lot of output. While running, fio will display the
178 status of the jobs created. An example of that would be:
180 Threads now running: 2 : [ww] [5.73% done]
182 The characters inside the square brackets denote the current status of
183 each thread. The possible values (in typical life cycle order) are:
187 P Thread setup, but not started.
188 C Thread created and running, but not doing anything yet
189 R Running, doing sequential reads.
190 r Running, doing random reads.
191 W Running, doing sequential writes.
192 w Running, doing random writes.
193 V Running, doing verification of written data.
194 E Thread exited, not reaped by main thread yet.
197 The other values are fairly self explanatory - number of thread currently
198 running and doing io, and the estimated completion percentage.
200 When fio is done (or interrupted by ctrl-c), it will show the data for
201 each thread, group of threads, and disks in that order. For each data
202 direction, the output looks like:
204 Client1 (g=0): err= 0:
205 write: io= 32MiB, bw= 666KiB/s, runt= 50320msec
206 slat (msec): min= 0, max= 136, avg= 0.03, dev= 1.92
207 clat (msec): min= 0, max= 631, avg=48.50, dev=86.82
208 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, dev=681.68
209 cpu : usr=1.49%, sys=0.25%, ctx=7969
211 The client number is printed, along with the group id and error of that
212 thread. Below is the io statistics, here for writes. In the order listed,
215 io= Number of megabytes io performed
216 bw= Average bandwidth rate
217 runt= The runtime of that thread
218 slat= Submission latency (avg being the average, dev being the
219 standard deviation). This is the time it took to submit
220 the io. For sync io, the slat is really the completion
221 latency, since queue/complete is one operation there.
222 clat= Completion latency. Same names as slat, this denotes the
223 time from submission to completion of the io pieces. For
224 sync io, clat will usually be equal (or very close) to 0,
225 as the time from submit to complete is basically just
226 CPU time (io has already been done, see slat explanation).
227 bw= Bandwidth. Same names as the xlat stats, but also includes
228 an approximate percentage of total aggregate bandwidth
229 this thread received in this group. This last value is
230 only really useful if the threads in this group are on the
231 same disk, since they are then competing for disk access.
232 cpu= CPU usage. User and system time, along with the number
233 of context switches this thread went through.
235 After each client has been listed, the group statistics are printed. They
238 Run status group 0 (all jobs):
239 READ: io=64MiB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
240 WRITE: io=64MiB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
242 For each data direction, it prints:
244 io= Number of megabytes io performed.
245 aggrb= Aggregate bandwidth of threads in this group.
246 minb= The minimum average bandwidth a thread saw.
247 maxb= The maximum average bandwidth a thread saw.
248 mint= The minimum runtime of a thread.
249 maxt= The maximum runtime of a thread.
251 And finally, the disk statistics are printed. They will look like this:
253 Disk stats (read/write):
254 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
256 Each value is printed for both reads and writes, with reads first. The
259 ios= Number of ios performed by all groups.
260 merge= Number of merges io the io scheduler.
261 ticks= Number of ticks we kept the disk busy.
262 io_queue= Total time spent in the disk queue.
263 util= The disk utilization. A value of 100% means we kept the disk
264 busy constantly, 50% would be a disk idling half of the time.