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