NAME¶
buffer - very fast reblocking program
SYNTAX¶
buffer [
-S size] [
-b blocks] [
-s size] [
-z
size] [
-m size] [
-p percentage] [
-u microseconds]
[
-B] [
-t] [
-Z] [
-i filename] [
-o filename]
[
-d]
OPTIONS¶
- -i filename
- Use the given file as the input file. The default is stdin.
- -o filename
- Use the given file as the output file. The default is stdout.
- -S size
- After every chunk of this size has been written, print out how much has
been written so far. Also prints the total throughput. By default this is
not set.
- -s size
- Size in bytes of each block. The default blocksize is 10k to match the
normal output of the tar(1) program.
- -z size
- Combines the -S and -s flags.
- -b blocks
- Number of blocks to allocate to shared memory circular buffer. Defaults to
the number required to fill up the shared memory requested.
- -m size
- Maximum size of the shared memory chunk to allocate for the circular
queue. Defaults to one megabyte.
- -p percentage
- Only start a write when the given percentage of the internal queue is
full. A percentage around 75 often proves best. Defaults to zero.
- -u microseconds
- After every write pause for this many microseconds. Defaults to zero.
(Surprisingly a small sleep, 100 usecs, after each write can greatly
enhance throughput on some drives.)
- -B
- Force each block written to be padded out to the blocksize. This is needed
by some tape and cartridge drives. Defaults to unpadded. This only affects
the last block written.
- -t
- On exiting print to stderr a brief message showing the total number of
bytes written.
- -Z
- If reading/writing directly to a character device (like a tape drive) then
after each gigabyte perform an lseek to the start of the file. Use this
flag with extreme care. It can only be used on devices where an lseek does
not rewind the tape but does reset the kernels position flags. It is used
to allow more than 2 gigabytes to be written.
- -d
- Print debug information to stderr during operation.
Sizes are a number with an optional trailing character. A 'b' multiplies the
size by 512, a 'k' by 1024 and an 'm' by a meg.
DESCRIPTION¶
Buffer reads from standard input reblocking to the given blocksize and
writes each block to standard output.
Internally
buffer is a pair of processes communicating via a large
circular queue held in shared memory. The reader process only has to block
when the queue is full and the writer process when the queue is empty.
Buffer is designed to try and keep the writer side continuously busy so
that it can stream when writing to tape drives. When used to write tapes with
an intervening network
buffer can result in a considerable increase in
throughput.
The default settings for
buffer are normally good enough. If you are a
heavy tape user then it is worth your while trying out various different
combinations of options. In particular running a
buffer at both ends of
the pipe can provide a substantial increase (see last example below).
EXAMPLES¶
$
buffer < /etc/termcap > /dev/rst8
$
tar cf - . | rsh somehost 'buffer > /dev/rst8'
$
dump fu - | rsh somehost 'buffer -s 16k > /dev/nrst8'
$
tar cf - . | buffer |
rsh somehost 'buffer -S 500K -p 75 > /dev/rst0'
BUFFER LIMITS¶
There are 2 limits in
buffer affecting the maximum total buffer size: The
maximum number of blocks (2048) and the maximum size of a block (512kB). This
results in a maximum total buffer size of 1GB.
Note that there is also a kernel limit for the maximum size of a shared memory
segment (used by
buffer internally) which is typically much lower than
the limits in
buffer. So if
buffer complains about not being
able to create a shared memory segment, this limit might have to be raised
using the command
sysctl kernel.shmmax=XXX (see also sysctl(1)
and
proc(5)).
BUGS¶
Internally, for printing purposes, buffer counts in terms of the number of
kilobytes output. If the blocksize you use is not a whole number of kilobytes
then the numbers printed will be inaccurate.
THANKS¶
Thanks to Kevin Twidle <kpt@doc.ic.ac.uk> for a lot of early suggestions
and patches to make it work with non-tar/dump tapes to exabyte drives.
Thanks to Andi Karrer <karrer@bernina.ethz.ch>, Rumi Zahir
<rumi@iis.ethz.ch> and Christoph Wicki <wicki@iis.ethz.ch> for
patches to make buffer work when trying to write single tape files of greater
than 2 gigabytes.
COPYRIGHT¶
Copyright (C) 1990, 1991 by Lee McLoughlin.
SEE ALSO¶
dd(1),
tar(1),
rsh(1)