8 5. Detailed list of parameters
13 1.0 Overview and history
14 ------------------------
15 fio was originally written to save me the hassle of writing special test
16 case programs when I wanted to test a specific workload, either for
17 performance reasons or to find/reproduce a bug. The process of writing
18 such a test app can be tiresome, especially if you have to do it often.
19 Hence I needed a tool that would be able to simulate a given io workload
20 without resorting to writing a tailored test case again and again.
22 A test work load is difficult to define, though. There can be any number
23 of processes or threads involved, and they can each be using their own
24 way of generating io. You could have someone dirtying large amounts of
25 memory in an memory mapped file, or maybe several threads issuing
26 reads using asynchronous io. fio needed to be flexible enough to
27 simulate both of these cases, and many more.
31 The first step in getting fio to simulate a desired io workload, is
32 writing a job file describing that specific setup. A job file may contain
33 any number of threads and/or files - the typical contents of the job file
34 is a global section defining shared parameters, and one or more job
35 sections describing the jobs involved. When run, fio parses this file
36 and sets everything up as described. If we break down a job from top to
37 bottom, it contains the following basic parameters:
39 IO type Defines the io pattern issued to the file(s).
40 We may only be reading sequentially from this
41 file(s), or we may be writing randomly. Or even
42 mixing reads and writes, sequentially or randomly.
44 Block size In how large chunks are we issuing io? This may be
45 a single value, or it may describe a range of
48 IO size How much data are we going to be reading/writing.
50 IO engine How do we issue io? We could be memory mapping the
51 file, we could be using regular read/write, we
52 could be using splice, async io, syslet, or even
55 IO depth If the io engine is async, how large a queuing
56 depth do we want to maintain?
58 IO type Should we be doing buffered io, or direct/raw io?
60 Num files How many files are we spreading the workload over.
62 Num threads How many threads or processes should we spread
65 The above are the basic parameters defined for a workload, in addition
66 there's a multitude of parameters that modify other aspects of how this
72 See the README file for command line parameters, there are only a few
75 Running fio is normally the easiest part - you just give it the job file
76 (or job files) as parameters:
80 and it will start doing what the job_file tells it to do. You can give
81 more than one job file on the command line, fio will serialize the running
82 of those files. Internally that is the same as using the 'stonewall'
83 parameter described the the parameter section.
85 If the job file contains only one job, you may as well just give the
86 parameters on the command line. The command line parameters are identical
87 to the job parameters, with a few extra that control global parameters
88 (see README). For example, for the job file parameter iodepth=2, the
89 mirror command line option would be --iodepth 2 or --iodepth=2. You can
90 also use the command line for giving more than one job entry. For each
91 --name option that fio sees, it will start a new job with that name.
92 Command line entries following a --name entry will apply to that job,
93 until there are no more entries or a new --name entry is seen. This is
94 similar to the job file options, where each option applies to the current
95 job until a new [] job entry is seen.
97 fio does not need to run as root, except if the files or devices specified
98 in the job section requires that. Some other options may also be restricted,
99 such as memory locking, io scheduler switching, and decreasing the nice value.
104 As previously described, fio accepts one or more job files describing
105 what it is supposed to do. The job file format is the classic ini file,
106 where the names enclosed in [] brackets define the job name. You are free
107 to use any ascii name you want, except 'global' which has special meaning.
108 A global section sets defaults for the jobs described in that file. A job
109 may override a global section parameter, and a job file may even have
110 several global sections if so desired. A job is only affected by a global
111 section residing above it. If the first character in a line is a ';' or a
112 '#', the entire line is discarded as a comment.
114 So lets look at a really simple job file that define to threads, each
115 randomly reading from a 128MiB file.
117 ; -- start job file --
128 As you can see, the job file sections themselves are empty as all the
129 described parameters are shared. As no filename= option is given, fio
130 makes up a filename for each of the jobs as it sees fit. On the command
131 line, this job would look as follows:
133 $ fio --name=global --rw=randread --size=128m --name=job1 --name=job2
136 Lets look at an example that have a number of processes writing randomly
139 ; -- start job file --
151 Here we have no global section, as we only have one job defined anyway.
152 We want to use async io here, with a depth of 4 for each file. We also
153 increased the buffer size used to 32KiB and define numjobs to 4 to
154 fork 4 identical jobs. The result is 4 processes each randomly writing
155 to their own 64MiB file. Instead of using the above job file, you could
156 have given the parameters on the command line. For this case, you would
159 $ fio --name=random-writers --ioengine=libaio --iodepth=4 --rw=randwrite --bs=32k --direct=0 --size=64m --numjobs=4
161 fio ships with a few example job files, you can also look there for
165 5.0 Detailed list of parameters
166 -------------------------------
168 This section describes in details each parameter associated with a job.
169 Some parameters take an option of a given type, such as an integer or
170 a string. The following types are used:
172 str String. This is a sequence of alpha characters.
173 int Integer. A whole number value, may be negative.
174 siint SI integer. A whole number value, which may contain a postfix
175 describing the base of the number. Accepted postfixes are k/m/g,
176 meaning kilo, mega, and giga. So if you want to specify 4096,
177 you could either write out '4096' or just give 4k. The postfixes
178 signify base 2 values, so 1024 is 1k and 1024k is 1m and so on.
179 bool Boolean. Usually parsed as an integer, however only defined for
180 true and false (1 and 0).
181 irange Integer range with postfix. Allows value range to be given, such
182 as 1024-4096. A colon may also be used as the seperator, eg
183 1k:4k. If the option allows two sets of ranges, they can be
184 specified with a ',' or '/' delimiter: 1k-4k/8k-32k. Also see
187 With the above in mind, here follows the complete list of fio job
190 name=str ASCII name of the job. This may be used to override the
191 name printed by fio for this job. Otherwise the job
192 name is used. On the command line this parameter has the
193 special purpose of also signaling the start of a new
196 description=str Text description of the job. Doesn't do anything except
197 dump this text description when this job is run. It's
200 directory=str Prefix filenames with this directory. Used to places files
201 in a different location than "./".
203 filename=str Fio normally makes up a filename based on the job name,
204 thread number, and file number. If you want to share
205 files between threads in a job or several jobs, specify
206 a filename for each of them to override the default. If
207 the ioengine used is 'net', the filename is the host and
208 port to connect to in the format of =host/port. If the
209 ioengine is file based, you can specify a number of files
210 by seperating the names with a ':' colon. So if you wanted
211 a job to open /dev/sda and /dev/sdb as the two working files,
212 you would use filename=/dev/sda:/dev/sdb
214 opendir=str Tell fio to recursively add any file it can find in this
215 directory and down the file system tree.
217 rw=str Type of io pattern. Accepted values are:
219 read Sequential reads
220 write Sequential writes
221 randwrite Random writes
222 randread Random reads
223 rw Sequential mixed reads and writes
224 randrw Random mixed reads and writes
226 For the mixed io types, the default is to split them 50/50.
227 For certain types of io the result may still be skewed a bit,
228 since the speed may be different.
230 randrepeat=bool For random IO workloads, seed the generator in a predictable
231 way so that results are repeatable across repetitions.
233 size=siint The total size of file io for this job. This may describe
234 the size of the single file the job uses, or it may be
235 divided between the number of files in the job. If the
236 file already exists, the file size will be adjusted to this
237 size if larger than the current file size. If this parameter
238 is not given and the file exists, the file size will be used.
240 filesize=siint Individual file sizes. May be a range, in which case fio
241 will select sizes for files at random within the given range
242 and limited to 'size' in total (if that is given). If not
243 given, each created file is the same size.
245 bs=siint The block size used for the io units. Defaults to 4k. Values
246 can be given for both read and writes. If a single siint is
247 given, it will apply to both. If a second siint is specified
248 after a comma, it will apply to writes only. In other words,
249 the format is either bs=read_and_write or bs=read,write.
250 bs=4k,8k will thus use 4k blocks for reads, and 8k blocks
251 for writes. If you only wish to set the write size, you
252 can do so by passing an empty read size - bs=,8k will set
253 8k for writes and leave the read default value.
255 bsrange=irange Instead of giving a single block size, specify a range
256 and fio will mix the issued io block sizes. The issued
257 io unit will always be a multiple of the minimum value
258 given (also see bs_unaligned). Applies to both reads and
259 writes, however a second range can be given after a comma.
262 bs_unaligned If this option is given, any byte size value within bsrange
263 may be used as a block range. This typically wont work with
264 direct IO, as that normally requires sector alignment.
266 nrfiles=int Number of files to use for this job. Defaults to 1.
268 openfiles=int Number of files to keep open at the same time. Defaults to
269 the same as nrfiles, can be set smaller to limit the number
272 file_service_type=str Defines how fio decides which file from a job to
273 service next. The following types are defined:
275 random Just choose a file at random.
277 roundrobin Round robin over open files. This
280 The string can have a number appended, indicating how
281 often to switch to a new file. So if option random:4 is
282 given, fio will switch to a new random file after 4 ios
285 ioengine=str Defines how the job issues io to the file. The following
288 sync Basic read(2) or write(2) io. lseek(2) is
289 used to position the io location.
291 libaio Linux native asynchronous io.
293 posixaio glibc posix asynchronous io.
295 mmap File is memory mapped and data copied
296 to/from using memcpy(3).
298 splice splice(2) is used to transfer the data and
299 vmsplice(2) to transfer data from user
302 syslet-rw Use the syslet system calls to make
303 regular read/write async.
305 sg SCSI generic sg v3 io. May either be
306 synchronous using the SG_IO ioctl, or if
307 the target is an sg character device
308 we use read(2) and write(2) for asynchronous
311 null Doesn't transfer any data, just pretends
312 to. This is mainly used to exercise fio
313 itself and for debugging/testing purposes.
315 net Transfer over the network to given host:port.
316 'filename' must be set appropriately to
317 filename=host/port regardless of send
318 or receive, if the latter only the port
321 cpu Doesn't transfer any data, but burns CPU
322 cycles according to the cpuload= and
323 cpucycle= options. Setting cpuload=85
324 will cause that job to do nothing but burn
327 external Prefix to specify loading an external
328 IO engine object file. Append the engine
329 filename, eg ioengine=external:/tmp/foo.o
330 to load ioengine foo.o in /tmp.
332 iodepth=int This defines how many io units to keep in flight against
333 the file. The default is 1 for each file defined in this
334 job, can be overridden with a larger value for higher
337 iodepth_batch=int This defines how many pieces of IO to submit at once.
338 It defaults to the same as iodepth, but can be set lower
341 iodepth_low=int The low water mark indicating when to start filling
342 the queue again. Defaults to the same as iodepth, meaning
343 that fio will attempt to keep the queue full at all times.
344 If iodepth is set to eg 16 and iodepth_low is set to 4, then
345 after fio has filled the queue of 16 requests, it will let
346 the depth drain down to 4 before starting to fill it again.
348 direct=bool If value is true, use non-buffered io. This is usually
351 buffered=bool If value is true, use buffered io. This is the opposite
352 of the 'direct' option. Defaults to true.
354 offset=siint Start io at the given offset in the file. The data before
355 the given offset will not be touched. This effectively
356 caps the file size at real_size - offset.
358 fsync=int If writing to a file, issue a sync of the dirty data
359 for every number of blocks given. For example, if you give
360 32 as a parameter, fio will sync the file for every 32
361 writes issued. If fio is using non-buffered io, we may
362 not sync the file. The exception is the sg io engine, which
363 synchronizes the disk cache anyway.
365 overwrite=bool If writing to a file, setup the file first and do overwrites.
367 end_fsync=bool If true, fsync file contents when the job exits.
369 fsync_on_close=bool If true, fio will fsync() a dirty file on close.
370 This differs from end_fsync in that it will happen on every
371 file close, not just at the end of the job.
373 rwmixcycle=int Value in milliseconds describing how often to switch between
374 reads and writes for a mixed workload. The default is
377 rwmixread=int How large a percentage of the mix should be reads.
379 rwmixwrite=int How large a percentage of the mix should be writes. If both
380 rwmixread and rwmixwrite is given and the values do not add
381 up to 100%, the latter of the two will be used to override
384 norandommap Normally fio will cover every block of the file when doing
385 random IO. If this option is given, fio will just get a
386 new random offset without looking at past io history. This
387 means that some blocks may not be read or written, and that
388 some blocks may be read/written more than once. This option
389 is mutually exclusive with verify= for that reason.
391 nice=int Run the job with the given nice value. See man nice(2).
393 prio=int Set the io priority value of this job. Linux limits us to
394 a positive value between 0 and 7, with 0 being the highest.
397 prioclass=int Set the io priority class. See man ionice(1).
399 thinktime=int Stall the job x microseconds after an io has completed before
400 issuing the next. May be used to simulate processing being
401 done by an application. See thinktime_blocks and
405 Only valid if thinktime is set - pretend to spend CPU time
406 doing something with the data received, before falling back
407 to sleeping for the rest of the period specified by
411 Only valid if thinktime is set - control how many blocks
412 to issue, before waiting 'thinktime' usecs. If not set,
413 defaults to 1 which will make fio wait 'thinktime' usecs
416 rate=int Cap the bandwidth used by this job to this number of KiB/sec.
418 ratemin=int Tell fio to do whatever it can to maintain at least this
421 ratecycle=int Average bandwidth for 'rate' and 'ratemin' over this number
424 cpumask=int Set the CPU affinity of this job. The parameter given is a
425 bitmask of allowed CPU's the job may run on. See man
426 sched_setaffinity(2).
428 startdelay=int Start this job the specified number of seconds after fio
429 has started. Only useful if the job file contains several
430 jobs, and you want to delay starting some jobs to a certain
433 runtime=int Tell fio to terminate processing after the specified number
434 of seconds. It can be quite hard to determine for how long
435 a specified job will run, so this parameter is handy to
436 cap the total runtime to a given time.
438 invalidate=bool Invalidate the buffer/page cache parts for this file prior
439 to starting io. Defaults to true.
441 sync=bool Use sync io for buffered writes. For the majority of the
442 io engines, this means using O_SYNC.
444 mem=str Fio can use various types of memory as the io unit buffer.
445 The allowed values are:
447 malloc Use memory from malloc(3) as the buffers.
449 shm Use shared memory as the buffers. Allocated
452 shmhuge Same as shm, but use huge pages as backing.
454 mmap Use mmap to allocate buffers. May either be
455 anonymous memory, or can be file backed if
456 a filename is given after the option. The
457 format is mem=mmap:/path/to/file.
459 mmaphuge Use a memory mapped huge file as the buffer
460 backing. Append filename after mmaphuge, ala
461 mem=mmaphuge:/hugetlbfs/file
463 The area allocated is a function of the maximum allowed
464 bs size for the job, multiplied by the io depth given. Note
465 that for shmhuge and mmaphuge to work, the system must have
466 free huge pages allocated. This can normally be checked
467 and set by reading/writing /proc/sys/vm/nr_hugepages on a
468 Linux system. Fio assumes a huge page is 4MiB in size. So
469 to calculate the number of huge pages you need for a given
470 job file, add up the io depth of all jobs (normally one unless
471 iodepth= is used) and multiply by the maximum bs set. Then
472 divide that number by the huge page size. You can see the
473 size of the huge pages in /proc/meminfo. If no huge pages
474 are allocated by having a non-zero number in nr_hugepages,
475 using mmaphuge or shmhuge will fail. Also see hugepage-size.
477 mmaphuge also needs to have hugetlbfs mounted and the file
478 location should point there. So if it's mounted in /huge,
479 you would use mem=mmaphuge:/huge/somefile.
482 Defines the size of a huge page. Must at least be equal
483 to the system setting, see /proc/meminfo. Defaults to 4MiB.
484 Should probably always be a multiple of megabytes, so using
485 hugepage-size=Xm is the preferred way to set this to avoid
486 setting a non-pow-2 bad value.
488 exitall When one job finishes, terminate the rest. The default is
489 to wait for each job to finish, sometimes that is not the
492 bwavgtime=int Average the calculated bandwidth over the given time. Value
493 is specified in milliseconds.
495 create_serialize=bool If true, serialize the file creating for the jobs.
496 This may be handy to avoid interleaving of data
497 files, which may greatly depend on the filesystem
498 used and even the number of processors in the system.
500 create_fsync=bool fsync the data file after creation. This is the
503 unlink=bool Unlink the job files when done. Not the default, as repeated
504 runs of that job would then waste time recreating the fileset
507 loops=int Run the specified number of iterations of this job. Used
508 to repeat the same workload a given number of times. Defaults
511 verify=str If writing to a file, fio can verify the file contents
512 after each iteration of the job. The allowed values are:
514 md5 Use an md5 sum of the data area and store
515 it in the header of each block.
517 crc32 Use a crc32 sum of the data area and store
518 it in the header of each block.
520 This option can be used for repeated burn-in tests of a
521 system to make sure that the written data is also
524 stonewall Wait for preceeding jobs in the job file to exit, before
525 starting this one. Can be used to insert serialization
526 points in the job file.
528 numjobs=int Create the specified number of clones of this job. May be
529 used to setup a larger number of threads/processes doing
530 the same thing. We regard that grouping of jobs as a
533 group_reporting If 'numjobs' is set, it may be interesting to display
534 statistics for the group as a whole instead of for each
535 individual job. This is especially true of 'numjobs' is
536 large, looking at individual thread/process output quickly
537 becomes unwieldy. If 'group_reporting' is specified, fio
538 will show the final report per-group instead of per-job.
540 thread fio defaults to forking jobs, however if this option is
541 given, fio will use pthread_create(3) to create threads
544 zonesize=siint Divide a file into zones of the specified size. See zoneskip.
546 zoneskip=siint Skip the specified number of bytes when zonesize data has
547 been read. The two zone options can be used to only do
548 io on zones of a file.
550 write_iolog=str Write the issued io patterns to the specified file. See
553 read_iolog=str Open an iolog with the specified file name and replay the
554 io patterns it contains. This can be used to store a
555 workload and replay it sometime later.
557 write_bw_log If given, write a bandwidth log of the jobs in this job
558 file. Can be used to store data of the bandwidth of the
559 jobs in their lifetime. The included fio_generate_plots
560 script uses gnuplot to turn these text files into nice
563 write_lat_log Same as write_bw_log, except that this option stores io
564 completion latencies instead.
566 lockmem=siint Pin down the specified amount of memory with mlock(2). Can
567 potentially be used instead of removing memory or booting
568 with less memory to simulate a smaller amount of memory.
570 exec_prerun=str Before running this job, issue the command specified
573 exec_postrun=str After the job completes, issue the command specified
576 ioscheduler=str Attempt to switch the device hosting the file to the specified
577 io scheduler before running.
579 cpuload=int If the job is a CPU cycle eater, attempt to use the specified
580 percentage of CPU cycles.
582 cpuchunks=int If the job is a CPU cycle eater, split the load into
583 cycles of the given time. In milliseconds.
586 6.0 Interpreting the output
587 ---------------------------
589 fio spits out a lot of output. While running, fio will display the
590 status of the jobs created. An example of that would be:
592 Threads: 1: [_r] [24.8% done] [ 13509/ 8334 kb/s] [eta 00h:01m:31s]
594 The characters inside the square brackets denote the current status of
595 each thread. The possible values (in typical life cycle order) are:
599 P Thread setup, but not started.
601 I Thread initialized, waiting.
602 R Running, doing sequential reads.
603 r Running, doing random reads.
604 W Running, doing sequential writes.
605 w Running, doing random writes.
606 M Running, doing mixed sequential reads/writes.
607 m Running, doing mixed random reads/writes.
608 F Running, currently waiting for fsync()
609 V Running, doing verification of written data.
610 E Thread exited, not reaped by main thread yet.
613 The other values are fairly self explanatory - number of threads
614 currently running and doing io, rate of io since last check, and the estimated
615 completion percentage and time for the running group. It's impossible to
616 estimate runtime of the following groups (if any).
618 When fio is done (or interrupted by ctrl-c), it will show the data for
619 each thread, group of threads, and disks in that order. For each data
620 direction, the output looks like:
622 Client1 (g=0): err= 0:
623 write: io= 32MiB, bw= 666KiB/s, runt= 50320msec
624 slat (msec): min= 0, max= 136, avg= 0.03, stdev= 1.92
625 clat (msec): min= 0, max= 631, avg=48.50, stdev=86.82
626 bw (KiB/s) : min= 0, max= 1196, per=51.00%, avg=664.02, stdev=681.68
627 cpu : usr=1.49%, sys=0.25%, ctx=7969
628 IO depths : 1=0.1%, 2=0.3%, 4=0.5%, 8=99.0%, 16=0.0%, 32=0.0%, >32=0.0%
629 lat (msec): 2=1.6%, 4=0.0%, 10=3.2%, 20=12.8%, 50=38.4%, 100=24.8%,
630 lat (msec): 250=15.2%, 500=0.0%, 750=0.0%, 1000=0.0%, >=2048=0.0%
632 The client number is printed, along with the group id and error of that
633 thread. Below is the io statistics, here for writes. In the order listed,
636 io= Number of megabytes io performed
637 bw= Average bandwidth rate
638 runt= The runtime of that thread
639 slat= Submission latency (avg being the average, dev being the
640 standard deviation). This is the time it took to submit
641 the io. For sync io, the slat is really the completion
642 latency, since queue/complete is one operation there.
643 clat= Completion latency. Same names as slat, this denotes the
644 time from submission to completion of the io pieces. For
645 sync io, clat will usually be equal (or very close) to 0,
646 as the time from submit to complete is basically just
647 CPU time (io has already been done, see slat explanation).
648 bw= Bandwidth. Same names as the xlat stats, but also includes
649 an approximate percentage of total aggregate bandwidth
650 this thread received in this group. This last value is
651 only really useful if the threads in this group are on the
652 same disk, since they are then competing for disk access.
653 cpu= CPU usage. User and system time, along with the number
654 of context switches this thread went through.
655 IO depths= The distribution of io depths over the job life time. The
656 numbers are divided into powers of 2, so for example the
657 16= entries includes depths up to that value but higher
658 than the previous entry. In other words, it covers the
660 IO latencies= The distribution of IO completion latencies. This is the
661 time from when IO leaves fio and when it gets completed.
662 The numbers follow the same pattern as the IO depths,
663 meaning that 2=1.6% means that 1.6% of the IO completed
664 within 2 msecs, 20=12.8% means that 12.8% of the IO
665 took more than 10 msecs, but less than (or equal to) 20 msecs.
667 After each client has been listed, the group statistics are printed. They
670 Run status group 0 (all jobs):
671 READ: io=64MiB, aggrb=22178, minb=11355, maxb=11814, mint=2840msec, maxt=2955msec
672 WRITE: io=64MiB, aggrb=1302, minb=666, maxb=669, mint=50093msec, maxt=50320msec
674 For each data direction, it prints:
676 io= Number of megabytes io performed.
677 aggrb= Aggregate bandwidth of threads in this group.
678 minb= The minimum average bandwidth a thread saw.
679 maxb= The maximum average bandwidth a thread saw.
680 mint= The smallest runtime of the threads in that group.
681 maxt= The longest runtime of the threads in that group.
683 And finally, the disk statistics are printed. They will look like this:
685 Disk stats (read/write):
686 sda: ios=16398/16511, merge=30/162, ticks=6853/819634, in_queue=826487, util=100.00%
688 Each value is printed for both reads and writes, with reads first. The
691 ios= Number of ios performed by all groups.
692 merge= Number of merges io the io scheduler.
693 ticks= Number of ticks we kept the disk busy.
694 io_queue= Total time spent in the disk queue.
695 util= The disk utilization. A value of 100% means we kept the disk
696 busy constantly, 50% would be a disk idling half of the time.
702 For scripted usage where you typically want to generate tables or graphs
703 of the results, fio can output the results in a semicolon separated format.
704 The format is one long line of values, such as:
706 client1;0;0;1906777;1090804;1790;0;0;0.000000;0.000000;0;0;0.000000;0.000000;929380;1152890;25.510151%;1078276.333333;128948.113404;0;0;0;0;0;0.000000;0.000000;0;0;0.000000;0.000000;0;0;0.000000%;0.000000;0.000000;100.000000%;0.000000%;324;100.0%;0.0%;0.0%;0.0%;0.0%;0.0%;0.0%;100.0%;0.0%;0.0%;0.0%;0.0%;0.0%
707 ;0.0%;0.0%;0.0%;0.0%;0.0%
709 Split up, the format is as follows:
711 jobname, groupid, error
713 KiB IO, bandwidth (KiB/sec), runtime (msec)
714 Submission latency: min, max, mean, deviation
715 Completion latency: min, max, mean, deviation
716 Bw: min, max, aggregate percentage of total, mean, deviation
718 KiB IO, bandwidth (KiB/sec), runtime (msec)
719 Submission latency: min, max, mean, deviation
720 Completion latency: min, max, mean, deviation
721 Bw: min, max, aggregate percentage of total, mean, deviation
722 CPU usage: user, system, context switches
723 IO depths: <=1, 2, 4, 8, 16, 32, >=64
724 IO latencies: <=2, 4, 10, 20, 50, 100, 250, 500, 750, 1000, >=2000