Mio(1) Mio(1)
NAME
Mio - Multi-stream streaming I/O test
SYNOPSIS
Mio [options] filename [filename...]
DESCRIPTION
Mio is a utility that can be used to measure the I/O performance, in
terms of bandwidth, of a system, I/O infrastructure, disk subsystem,
disk device and/or file system. Mio uses pthreads to asynchronously
queue requests to a stream or streams.
Each file name specified on the command line corresponds to a stream.
The -q option defaults to 2 as described below, and specifies the num-
ber of asynchronous read or write requests that are queued to each
stream. A file name can represent a regular file, or a block or char-
acter device (see NOTES). By default, Mio will issue read requests,
with a -w option required for writes.
Upon completion, Mio will print a summary of I/O performance for each
stream as well as the aggregate performance off all streams. Because
the aggregate performance is the performance of all streams over the
test run time, it may not reflect the sum of the individual streams.
OPTIONS
-? Display usage.
-b buffer_size
Specify the I/O buffer size to use. This value represents the
size of each I/O request in bytes. The default buffer size val-
ue is 1048576 bytes. Optionally, a single letter suffix can
represent bytes in units. Mio single letter unit suffixes are
as follows.
k KB 10^3 1,000
m MB 10^6 1,000,000
b GB 10^9 1,000,000,000
K KiB 2^10 1,024
M MiB 2^20 1,048,576
B GiB 2^30 1,073,741,824
-B Do buffered I/O. By default, Mio will open files for direct
I/O. See NOTES below for more information on direct I/O.
-c Create the file(s). This is useful and necessary for regular
files, as they must be created before they can be read or writ-
ten. The -c option is valid only when specified with the -w op-
tion.
-d[dd] Run in debug mode. The more "d's" specified, the more debug in-
formation is printed.
-e csv_file
Write results to the specified .csv file; one that can be read
by Microsoft Excel or other spreadsheet application. Data is ap-
pended to the specified csv file which must exist. See the -E
option.
-E csv_file
Create and write results to the specified .csv file. A .csv
file, (comma separated value) is a file that can be consumed by
a spreadsheet application like Microsoft Excel. The maximum
number of streams that can be recorded in a .csv file is limited
to 16 unless the -u option is specified.
-f Fsync on close. The default is no fsync.
-I file_offset_increment
Set the file offset increment for each I/O. For example, when
set to zero, we repetitively do I/O to the same file location.
The default is to use the buffer size. An Mio single letter
unit suffix may be specified.
-n nios
Specifies the number of I/O requests. The default is 1024.
-O file_offset
Start the I/O at the given file offset. An Mio single letter
unit suffix may be specified. The default starting file offset
for a regular file is zero. The default starting file offset
for a block device is two times the I/O buffer size. This al-
lows writing to a block device without destroying StorNext la-
bels (writing to the block device will, however, destroy the
file system).
-p Preallocate the file using the StorNext CvApi_VerifyAlloc API
call. The file size will correspond to the number of I/Os times
the buffer size. This option is supported only for regular
files on a StorNext file system.
-q queue_depth
Queue the specified number of I/O requests on each stream. The
default is 2. The maximum is currently 256.
-r rtios
Set realtime mode by setting the number of I/O's per second to
the specified value. Realtime mode is available only for regu-
lar files on the StorNext file system. An Mio single letter
unit suffix may be specified.
-R Record I/O response times. Mio will record the I/O response time
for each I/O request and increment a counter in the correspond-
ing response time bucket. I/O response time buckets are then
printed for each stream along with the transfer rate summary.
-S seconds
Run the test for the maximum number of seconds specified. This
only has an effect as a maximum, in that, if the test has not
completed the specified number of requests after the specified
time.
-t total_data
Terminate the test after the specified amount of data has been
moved rather than a given number of requests. This option is
not compatible with the -n option. An Mio single letter unit
suffix may be specified.
-T tag string
Specify a tag string that will be written to the output. An ad-
ditional line containing the specified tag string will be writ-
ten in the test summary. If the -E option is specified, the tag
string will also be written to the .csv file.
-u Specify that the unlimited streams format be used for csv files.
Each stream's data is written to a new line in the file allowing
for greater than 16 streams.
-w Specify that writes are done. A regular file must exist, or the
-c option must be specified to create the file. The default is
to read.
-W write_mask
The specified write mask will result in writes on streams where
bits are set and reads on streams where bits are clear. The
write mask may be specified in hexadecimal, octal, or decimal
but is limited to 64 bits, hence 64 streams. The -W option is
incompatible with the -c option and all files must exist.
EXAMPLES
Create and issue the default number of writes to four files/streams:
per1-# Mio -cw /jhb/f0 /jhb/f1 /jhb/f2 /jhb/f3
Mio: Timing 4 stream(s) of 1024 x 1M direct writes queued 2 deep
stream[0]: f0: write 1073.74 MBytes @ 256.44 MBytes/Second
stream[1]: f1: write 1073.74 MBytes @ 228.43 MBytes/Second
stream[2]: f2: write 1073.74 MBytes @ 265.04 MBytes/Second
stream[3]: f3: write 1073.74 MBytes @ 242.18 MBytes/Second
Mio: Aggregate: 4294.97 Mbytes @ 913.72 MBytes/Second
Issue the default number of reads to the four previously created files:
per1-# Mio /jhb/f0 /jhb/f1 /jhb/f2 /jhb/f3
Mio: Timing 4 stream(s) of 1024 x 1M direct reads queued 2 deep
stream[0]: f0: read 1073.74 MBytes @ 1329.16 MBytes/Second
stream[1]: f1: read 1073.74 MBytes @ 1334.83 MBytes/Second
stream[2]: f2: read 1073.74 MBytes @ 1316.28 MBytes/Second
stream[3]: f3: read 1073.74 MBytes @ 1329.92 MBytes/Second
Mio: Aggregate: 4294.97 Mbytes @ 5265.12 MBytes/Second
Issue the default number of reads to two block devices:
per1-# Mio /dev/sdd /dev/sdf
Mio: Timing 2 stream(s) of 1024 x 1M direct reads queued 2 deep
stream[0]: sdd: read 1073.74 MBytes @ 408.38 MBytes/Second
stream[1]: sdf: read 1073.74 MBytes @ 407.78 MBytes/Second
Mio: Aggregate: 2147.48 Mbytes @ 815.57 MBytes/Second
Create and do 10000 writes to four files via four streams:
per1-# Mio -cw -n10000 /jhb/f0 /jhb/f1 /jhb/f2 /jhb/f3
Mio: Timing 4 stream(s) of 10000 x 1M direct writes queued 2 deep
stream[0]: f0: write 10485.76 MBytes @ 488.97 MBytes/Second
stream[1]: f1: write 10485.76 MBytes @ 448.84 MBytes/Second
stream[2]: f2: write 10485.76 MBytes @ 458.36 MBytes/Second
stream[3]: f3: write 10485.76 MBytes @ 497.57 MBytes/Second
Mio: Aggregate: 41943.04 Mbytes @ 1795.35 MBytes/Second
Issue 10000 reads, queued 8 deep to four streams:
per1-# Mio -q8 -n10000 /jhb/f0 /jhb/f1 /jhb/f2 /jhb/f3
Mio: Timing 4 stream(s) of 10000 x 1M direct reads queued 8 deep
stream[0]: f0: read 10485.76 MBytes @ 898.16 MBytes/Second
stream[1]: f1: read 10485.76 MBytes @ 1075.04 MBytes/Second
stream[2]: f2: read 10485.76 MBytes @ 1050.13 MBytes/Second
stream[3]: f3: read 10485.76 MBytes @ 869.96 MBytes/Second
Mio: Aggregate: 41943.04 Mbytes @ 3479.83 MBytes/Second
Issue 240 x 12746752 reads to four streams and "tag" the summary:
per1-# Mio -b12746752 -n240 -T full_aperture_2K /jhb/f0 /jhb/f1 /jhb/f2 /jhb/f3
Mio: full_aperture_2K
Mio: Timing 4 stream(s) of 240 x 12448K direct reads queued 2 deep
stream[0]: f0: read 3059.22 MBytes @ 884.50 MBytes/Second
stream[1]: f1: read 3059.22 MBytes @ 726.70 MBytes/Second
stream[2]: f2: read 3059.22 MBytes @ 808.67 MBytes/Second
stream[3]: f3: read 3059.22 MBytes @ 656.74 MBytes/Second
Mio: Aggregate: 12236.88 Mbytes @ 2626.95 MBytes/Second
Issue 10000 reads to four streams and report the response times:
per1-# Mio -R -n10000 /jhb/f0 /jhb/f1 /jhb/f2 /jhb/f3
Mio: Timing 4 stream(s) of 10000 x 1M direct reads queued 2 deep
stream[0]: f0: read 10485.76 MBytes @ 536.02 MBytes/Second
stream[1]: f1: read 10485.76 MBytes @ 688.96 MBytes/Second
stream[2]: f2: read 10485.76 MBytes @ 856.74 MBytes/Second
stream[3]: f3: read 10485.76 MBytes @ 841.03 MBytes/Second
Mio: Aggregate: 41943.04 Mbytes @ 2144.08 MBytes/Second
I/O response time buckets: 0ms 5ms 10ms 20ms 50ms 100ms 500ms
stream[0]: f0: 7632 1459 768 137 4 0 0
stream[1]: f1: 8570 928 395 103 4 0 0
stream[2]: f2: 9240 491 197 68 4 0 0
stream[3]: f3: 9196 494 234 76 0 0 0
NOTES
Writing to a device that contains data can result in data loss. Do not
write to drives that contain any useful information.
Character devices - Some platforms have character device representa-
tions of disk devices and/or disk partitions that support read/write
and some do not.
Direct I/O - Direct I/O is not supported on all platforms, although it
is pretty well tested on Linux and Windows. Direct I/O support for
specific files on other platforms is schizophrenic at best, and may not
be supported on a particular file system, block device, character spe-
cial device, or without appropriate mount options.
Windows - Mio on Windows can be used from either a cygwin environment
or a DOS shell environment. File names are specified using the drive
letter format. Use of a leading slash is recommended to avoid inter-
pretation of the path based on the last relative location in that
drive. A Windows physical drive is specified using the \\.\ notation.
For example, PhysicalDrive9 is specified as \\.\PhysicalDrive9.
The Windows cygwin environment requires escaping the backslash result-
ing in 4 backslashes, period, 2 backslashes. Failure to escape the
backslash cygwin can cause writing to the root directory as the name is
changed to \.PhysicalDrive9.
FILES
/System/Library/Filesystems/acfs.fs/Contents/bin/Mio
SEE ALSO
cvfs(8)
StorNext File System January 2016 Mio(1)
Mac OS X 10.12.3 - Generated Sat Feb 4 18:03:25 CST 2017