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/
28 -b block size in KiB for each io
29 -t <sec> Runtime in seconds
30 -r For random io, sequence must be repeatable
31 -R <on> If one thread fails to meet rate, quit all
32 -o <on> Use direct IO is 1, buffered if 0
33 -l Generate per-job latency logs
34 -w Generate per-job bandwidth logs
35 -f <file> Read <file> for job descriptions
36 -v Print version information and exit
38 The <jobs> format is as follows:
40 directory=x Use 'x' as the top level directory for storing files
41 rw=x 'x' may be: read, randread, write, or randwrite
42 size=x Set file size to x bytes (x string can include k/m/g)
43 ioengine=x 'x' may be: aio/libaio/linuxaio for Linux aio,
44 posixaio for POSIX aio, sync for regular read/write io,
45 mmap for mmap'ed io, or sgio for direct SG_IO io. The
46 latter only works on Linux on SCSI (or SCSI-like
47 devices, such as usb-storage or sata/libata driven)
49 iodepth=x For async io, allow 'x' ios in flight
50 overwrite=x If 'x', layout a write file first.
51 prio=x Run io at prio X, 0-7 is the kernel allowed range
52 prioclass=x Run io at prio class X
53 bs=x Use 'x' for thread blocksize. May include k/m postfix.
54 bsrange=x-y Mix thread block sizes randomly between x and y. May
55 also include k/m postfix.
56 direct=x 1 for direct IO, 0 for buffered IO
57 thinktime=x "Think" x usec after each io
58 rate=x Throttle rate to x KiB/sec
59 ratemin=x Quit if rate of x KiB/sec can't be met
60 ratecycle=x ratemin averaged over x msecs
61 cpumask=x Only allow job to run on CPUs defined by mask.
62 fsync=x If writing, fsync after every x blocks have been written
63 startdelay=x Start this thread x seconds after startup
64 timeout=x Terminate x seconds after startup
65 offset=x Start io at offset x (x string can include k/m/g)
66 invalidate=x Invalidate page cache for file prior to doing io
67 sync=x Use sync writes if x and writing
68 mem=x If x == malloc, use malloc for buffers. If x == shm,
69 use shm for buffers. If x == mmap, use anon mmap.
70 exitall When one thread quits, terminate the others
71 bwavgtime=x Average bandwidth stats over an x msec window.
72 create_serialize=x If 'x', serialize file creation.
73 create_fsync=x If 'x', run fsync() after file creation.
74 loops=x Run the job 'x' number of times.
75 verify=x If 'x' == md5, use md5 for verifies. If 'x' == crc32,
76 use crc32 for verifies. md5 is 'safer', but crc32 is
77 a lot faster. Only makes sense for writing to a file.
78 stonewall Wait for preceeding jobs to end before running.
79 numjobs=x Create 'x' similar entries for this job
80 thread Use pthreads instead of forked jobs
82 zoneskip=y Zone options must be paired. If given, the job
83 will skip y bytes for every x read/written. This
84 can be used to gauge hard drive speed over the entire
85 platter, without reading everything. Both x/y can
89 Examples using a job file
90 -------------------------
92 A sample job file doing the same as above would look like this:
102 And fio would be invoked as:
104 $ fio -o1 -s -f file_with_above
106 The second example would look like this:
121 And fio would be invoked as:
123 $ fio -o0 -s -b4096 -f file_with_above
125 'global' is a reserved keyword. When used as the filename, it sets the
126 default options for the threads following that section. It is possible
127 to have more than one global section in the file, as it only affects
130 Also see the examples/ dir for sample job files.
133 Interpreting the output
134 -----------------------
136 fio spits out a lot of output. While running, fio will display the
137 status of the jobs created. An example of that would be:
139 Threads now running: 2 : [ww] [5.73% done]
141 The characters inside the square brackets denote the current status of
142 each thread. The possible values (in typical life cycle order) are:
146 P Thread setup, but not started.
147 C Thread created and running, but not doing anything yet
148 R Running, doing sequential reads.
149 r Running, doing random reads.
150 W Running, doing sequential writes.
151 w Running, doing random writes.
152 V Running, doing verification of written data.
153 E Thread exited, not reaped by main thread yet.
156 The other values are fairly self explanatory - number of thread currently
157 running and doing io, and the estimated completion percentage.
159 When fio is done (or interrupted by ctrl-c), it will show the data for
160 each thread, group of threads, and disks in that order. For each data
161 direction, the output looks like:
163 Client1 (g=0): err= 0:
164 write: io= 32MiB, bw= 666KiB/s, runt= 50320msec
165 slat (msec): min= 0, max= 136, avg= 0.03, dev= 1.92
166 clat (msec): min= 0, max= 631, avg=48.50, dev=86.82
167 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, dev=681.68
168 cpu : usr=1.49%, sys=0.25%, ctx=7969
170 The client number is printed, along with the group id and error of that
171 thread. Below is the io statistics, here for writes. In the order listed,
174 io= Number of megabytes io performed
175 bw= Average bandwidth rate
176 runt= The runtime of that thread
177 slat= Submission latency (avg being the average, dev being the
178 standard deviation). This is the time it took to submit
179 the io. For sync io, the slat is really the completion
180 latency, since queue/complete is one operation there.
181 clat= Completion latency. Same names as slat, this denotes the
182 time from submission to completion of the io pieces. For
183 sync io, clat will usually be equal (or very close) to 0,
184 as the time from submit to complete is basically just
185 CPU time (io has already been done, see slat explanation).
186 bw= Bandwidth. Same names as the xlat stats, but also includes
187 an approximate percentage of total aggregate bandwidth
188 this thread received in this group. This last value is
189 only really useful if the threads in this group are on the
190 same disk, since they are then competing for disk access.
191 cpu= CPU usage. User and system time, along with the number
192 of context switches this thread went through.
194 After each client has been listed, the group statistics are printed. They
197 Run status group 0 (all jobs):
198 READ: io=64MiB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
199 WRITE: io=64MiB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
201 For each data direction, it prints:
203 io= Number of megabytes io performed.
204 aggrb= Aggregate bandwidth of threads in this group.
205 minb= The minimum average bandwidth a thread saw.
206 maxb= The maximum average bandwidth a thread saw.
207 mint= The minimum runtime of a thread.
208 maxt= The maximum runtime of a thread.
210 And finally, the disk statistics are printed. They will look like this:
212 Disk stats (read/write):
213 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
215 Each value is printed for both reads and writes, with reads first. The
218 ios= Number of ios performed by all groups.
219 merge= Number of merges io the io scheduler.
220 ticks= Number of ticks we kept the disk busy.
221 io_queue= Total time spent in the disk queue.
222 util= The disk utilization. A value of 100% means we kept the disk
223 busy constantly, 50% would be a disk idling half of the time.