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
37 -v Print version information and exit
39 The <jobs> format is as follows:
41 directory=x Use 'x' as the top level directory for storing files
42 rw=x 'x' may be: read, randread, write, randwrite,
43 rw (read-write mix), randrw (read-write random mix)
44 rwmixcycle=x Base cycle for switching between read and write
46 rwmixread=x 'x' percentage of rw mix ios will be reads. If
47 rwmixwrite is also given, the last of the two will
48 be used if they don't add up to 100%.
49 rwmixwrite=x 'x' percentage of rw mix ios will be writes. See
51 size=x Set file size to x bytes (x string can include k/m/g)
52 ioengine=x 'x' may be: aio/libaio/linuxaio for Linux aio,
53 posixaio for POSIX aio, sync for regular read/write io,
54 mmap for mmap'ed io, splice for using splice/vmsplice,
55 or sgio for direct SG_IO io. The latter only works on
56 Linux on SCSI (or SCSI-like devices, such as
57 usb-storage or sata/libata driven) devices.
58 iodepth=x For async io, allow 'x' ios in flight
59 overwrite=x If 'x', layout a write file first.
60 prio=x Run io at prio X, 0-7 is the kernel allowed range
61 prioclass=x Run io at prio class X
62 bs=x Use 'x' for thread blocksize. May include k/m postfix.
63 bsrange=x-y Mix thread block sizes randomly between x and y. May
64 also include k/m postfix.
65 direct=x 1 for direct IO, 0 for buffered IO
66 thinktime=x "Think" x usec after each io
67 rate=x Throttle rate to x KiB/sec
68 ratemin=x Quit if rate of x KiB/sec can't be met
69 ratecycle=x ratemin averaged over x msecs
70 cpumask=x Only allow job to run on CPUs defined by mask.
71 fsync=x If writing, fsync after every x blocks have been written
72 startdelay=x Start this thread x seconds after startup
73 timeout=x Terminate x seconds after startup
74 offset=x Start io at offset x (x string can include k/m/g)
75 invalidate=x Invalidate page cache for file prior to doing io
76 sync=x Use sync writes if x and writing
77 mem=x If x == malloc, use malloc for buffers. If x == shm,
78 use shm for buffers. If x == mmap, use anon mmap.
79 exitall When one thread quits, terminate the others
80 bwavgtime=x Average bandwidth stats over an x msec window.
81 create_serialize=x If 'x', serialize file creation.
82 create_fsync=x If 'x', run fsync() after file creation.
83 end_fsync=x If 'x', run fsync() after end-of-job.
84 loops=x Run the job 'x' number of times.
85 verify=x If 'x' == md5, use md5 for verifies. If 'x' == crc32,
86 use crc32 for verifies. md5 is 'safer', but crc32 is
87 a lot faster. Only makes sense for writing to a file.
88 stonewall Wait for preceeding jobs to end before running.
89 numjobs=x Create 'x' similar entries for this job
90 thread Use pthreads instead of forked jobs
92 zoneskip=y Zone options must be paired. If given, the job
93 will skip y bytes for every x read/written. This
94 can be used to gauge hard drive speed over the entire
95 platter, without reading everything. Both x/y can
97 iolog=x Open and read io pattern from file 'x'. The file must
98 contain one io action per line in the following format:
100 where with rw=0/1 for read/write, and the offset
101 and length entries being in bytes.
102 write_iolog=x Write an iolog to file 'x' in the same format as iolog.
103 The iolog options are exclusive, if both given the
104 read iolog will be performed.
105 lockmem=x Lock down x amount of memory on the machine, to
106 simulate a machine with less memory available. x can
107 include k/m/g suffix.
108 nice=x Run job at given nice value.
110 Examples using a job file
111 -------------------------
113 A sample job file doing the same as above would look like this:
123 And fio would be invoked as:
125 $ fio -o1 -s -f file_with_above
127 The second example would look like this:
142 And fio would be invoked as:
144 $ fio -o0 -s -b4096 -f file_with_above
146 'global' is a reserved keyword. When used as the filename, it sets the
147 default options for the threads following that section. It is possible
148 to have more than one global section in the file, as it only affects
151 Also see the examples/ dir for sample job files.
154 Interpreting the output
155 -----------------------
157 fio spits out a lot of output. While running, fio will display the
158 status of the jobs created. An example of that would be:
160 Threads now running: 2 : [ww] [5.73% done]
162 The characters inside the square brackets denote the current status of
163 each thread. The possible values (in typical life cycle order) are:
167 P Thread setup, but not started.
168 C Thread created and running, but not doing anything yet
169 R Running, doing sequential reads.
170 r Running, doing random reads.
171 W Running, doing sequential writes.
172 w Running, doing random writes.
173 V Running, doing verification of written data.
174 E Thread exited, not reaped by main thread yet.
177 The other values are fairly self explanatory - number of thread currently
178 running and doing io, and the estimated completion percentage.
180 When fio is done (or interrupted by ctrl-c), it will show the data for
181 each thread, group of threads, and disks in that order. For each data
182 direction, the output looks like:
184 Client1 (g=0): err= 0:
185 write: io= 32MiB, bw= 666KiB/s, runt= 50320msec
186 slat (msec): min= 0, max= 136, avg= 0.03, dev= 1.92
187 clat (msec): min= 0, max= 631, avg=48.50, dev=86.82
188 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, dev=681.68
189 cpu : usr=1.49%, sys=0.25%, ctx=7969
191 The client number is printed, along with the group id and error of that
192 thread. Below is the io statistics, here for writes. In the order listed,
195 io= Number of megabytes io performed
196 bw= Average bandwidth rate
197 runt= The runtime of that thread
198 slat= Submission latency (avg being the average, dev being the
199 standard deviation). This is the time it took to submit
200 the io. For sync io, the slat is really the completion
201 latency, since queue/complete is one operation there.
202 clat= Completion latency. Same names as slat, this denotes the
203 time from submission to completion of the io pieces. For
204 sync io, clat will usually be equal (or very close) to 0,
205 as the time from submit to complete is basically just
206 CPU time (io has already been done, see slat explanation).
207 bw= Bandwidth. Same names as the xlat stats, but also includes
208 an approximate percentage of total aggregate bandwidth
209 this thread received in this group. This last value is
210 only really useful if the threads in this group are on the
211 same disk, since they are then competing for disk access.
212 cpu= CPU usage. User and system time, along with the number
213 of context switches this thread went through.
215 After each client has been listed, the group statistics are printed. They
218 Run status group 0 (all jobs):
219 READ: io=64MiB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
220 WRITE: io=64MiB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
222 For each data direction, it prints:
224 io= Number of megabytes io performed.
225 aggrb= Aggregate bandwidth of threads in this group.
226 minb= The minimum average bandwidth a thread saw.
227 maxb= The maximum average bandwidth a thread saw.
228 mint= The minimum runtime of a thread.
229 maxt= The maximum runtime of a thread.
231 And finally, the disk statistics are printed. They will look like this:
233 Disk stats (read/write):
234 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
236 Each value is printed for both reads and writes, with reads first. The
239 ios= Number of ios performed by all groups.
240 merge= Number of merges io the io scheduler.
241 ticks= Number of ticks we kept the disk busy.
242 io_queue= Total time spent in the disk queue.
243 util= The disk utilization. A value of 100% means we kept the disk
244 busy constantly, 50% would be a disk idling half of the time.