mirror of
https://github.com/10h30/lookbusy.git
synced 2026-05-12 15:21:20 +09:00
Copy from devin.com/lookbusy
This commit is contained in:
flow2000
+215
@@ -0,0 +1,215 @@
|
||||
." $Id$
|
||||
.TH lookbusy
|
||||
.SH NAME
|
||||
lookbusy \- a synthetic load generator
|
||||
|
||||
.SH SYNOPSIS
|
||||
\fBlookbusy\fR [OPTIONS]
|
||||
|
||||
.SH DESCRIPTION
|
||||
|
||||
lookbusy generates synthetic CPU, memory and disk access loads on a host. CPU
|
||||
load is induced by simple arithmetic looping (with as little memory bandwidth
|
||||
consumption as feasible) alternating with periods of sleeping in an attempt to
|
||||
generate the degree of utilization selected. Memory load is induced by
|
||||
allocating a buffer of a controllable size, then steadily stirring it to keep
|
||||
the pages active from the VM standpoint. Disk load is induced through
|
||||
creation of one or more files, and copying blocks of it between two moving
|
||||
positions.
|
||||
|
||||
.SH OPTIONS
|
||||
.TP
|
||||
\-h, \-\-help
|
||||
|
||||
Display commandline help.
|
||||
|
||||
.TP
|
||||
\-V, \-\-version
|
||||
|
||||
Display the lookbusy version and copyright information.
|
||||
|
||||
.TP
|
||||
\-v, \-\-verbose
|
||||
|
||||
Produce verbose output; may be specified more than once for additional noise.
|
||||
|
||||
.TP
|
||||
\-q, \-\-quiet
|
||||
|
||||
Be quiet; produce error output only.
|
||||
|
||||
.TP
|
||||
\-c \fIutil\fR[\-\fIhigh_util\fR], \-\-cpu\-util \fIutil\fR[\-\fIhigh_util\fR]
|
||||
|
||||
Attempt to use between \fIutil\fR and \fIhigh_util\fR percent of time on each
|
||||
CPU. For fixed usage mode (see \fB\-\-cpu\-mode\fR), only \fIutil\fR should
|
||||
be given. For curved usage mode, a range should be given. If specified,
|
||||
\fIhigh_util\fR should be greater than \fIutil\fR, and both values must be
|
||||
between 0 and 100 inclusive.
|
||||
|
||||
Prior to starting, a short benchmark will be attempted to establish a
|
||||
spin/sleep ratio needed; this estimate assumes that all CPUs are of equal
|
||||
speed, and that at least one CPU is unutilized during the test. The default
|
||||
is 50%.
|
||||
|
||||
.TP
|
||||
\-n \fIn\fR, \-\-ncpus \fIn\fR
|
||||
|
||||
Specify explicitly that \fIn\fR CPUs are to be kept busy, instead of
|
||||
attempting to auto-detect how many physical CPUs are involved.
|
||||
|
||||
.TP
|
||||
\-r \fImode\fR, \-\-cpu\-mode \fImode\fR
|
||||
|
||||
Select a CPU utilization mode. \fImode\fR must be one of \fBfixed\fR
|
||||
or \fBcurve\fR. The default is \fBfixed\fR.
|
||||
|
||||
In \fBfixed\fR mode, lookbusy will attempt to maintain a constant CPU usage at
|
||||
the level specified (see \fI\-\-cpu\-util\fR).
|
||||
|
||||
In \fBcurve\fR mode, CPU usage will vary between the range specified (again,
|
||||
see \fI\-\-cpu\-util\fR) according to a repeating curve function having a
|
||||
period specified by \fI\-\-cpu\-curve\-period\fR and a peak at the offset
|
||||
within that period specified by \fI\-\-cpu\-curve\-peak\fR. The default curve
|
||||
period is 24 hours long, with a peak at local midnight (or midnight GMT, if
|
||||
\-\-utc is given). As of version 1.0, this curve is simply a cosine function,
|
||||
though this may change in future releases.
|
||||
|
||||
.TP
|
||||
\-P \fIinterval\fR[\fIunit\fR], \-\-cpu\-curve\-period \fIinterval\fR[\fIunit\fR]
|
||||
|
||||
When using the \fBcurve\fR CPU usage mode, this option specifies the
|
||||
repetition period (frequency). The utilization curve will return to its peak
|
||||
once per interval. The interval is be specified as an integer, optionally
|
||||
followed by one of \fBd\fR (days), \fBh\fR (hours), \fBm\fR (minutes) or
|
||||
\fBs\fR (seconds). For example, \fB12h\fR would specify a 12-hour utilization
|
||||
cycle. If no units are specified, the value is interpreted as a number of
|
||||
seconds.
|
||||
|
||||
.TP
|
||||
\-p \fIoffset\fR[\fIunit\fR], \-\-cpu\-curve\-peak \fIinterval\fR[\fIunit\fR]
|
||||
|
||||
When using the \fBcurve\fR CPU mode, this option specifies the offset of the
|
||||
peak within the usage cycle. As with \fB\-\-cpu\-curve\-period\fR above,
|
||||
\fIoffset\fR is given as an integer, optionally followed by \fBd\fR, \fBh\fR,
|
||||
\fBm\fR or \fBs\fR. The default is chosen so as to place the peak at midnight
|
||||
local time (or UTC, if \-\-utc is given).
|
||||
|
||||
.TP
|
||||
\-u, \-\-utc
|
||||
|
||||
When computing time-relative utilization curves, make these computations for
|
||||
the UTC/GMT timezone. If not specified, the host timezone is used.
|
||||
|
||||
.TP
|
||||
\-m \fIutil\fR, \-\-mem-util \fIutil\fR
|
||||
|
||||
Keep \fIutil\fR mebibytes (1024^2 bytes) of memory utilized. This memory
|
||||
will be allocated, then continually stirred to ensure it is actually
|
||||
allocated in the VM system and impose pressure to keep it resident. The
|
||||
default is 0 (disabled).
|
||||
|
||||
.TP
|
||||
\-M \fIinterval\fR, \-\-mem\-sleep \fIinterval\fIR
|
||||
|
||||
Sleep \fIinterval\fR milliseconds after each memory stir iteration. The
|
||||
iteration reads and writes \fBPAGE_SIZE\fR bytes each from independently
|
||||
cycling positions in the working buffer. The default is 1ms, which will
|
||||
produce very little visible load on modern hardware, but will cover a 1GB
|
||||
utilization buffer in roughly 4 minutes.
|
||||
|
||||
.TP
|
||||
\-d \fIsize\fR[\fIunit\fR], \-\-disk\-util \fIsize\fR[\fIunit\fR]
|
||||
|
||||
Generate disk activity by creating and updating a buffer of the given size.
|
||||
The integer value \fIsize\fR may optionally be followed by one of \fBb\fR
|
||||
(bytes), \fBkb\fR (kilobytes), \fBmb\fR (megabytes), \fBgb\fR (gigabytes) or
|
||||
\fBtb\fR (terabytes). The default is 0 (disk churn disabled).
|
||||
|
||||
.TP
|
||||
\-D \fIinterval\fR, \-\-disk\-sleep \fIinterval\fR
|
||||
|
||||
Sleep \fIinterval\fR milliseconds after each iteration of the disk churn loop.
|
||||
Each iteration will read and write a block of data (see \fB\-b\fR) from two
|
||||
independently cycling positions in a scratch file. See also \fB\-f\fR. The
|
||||
default is 100ms.
|
||||
|
||||
.TP
|
||||
\-b \fIblocksize\fR[\fIunit\fR], \-\-disk\-block\-size \fIblocksize\fR[\fIunit\fR]
|
||||
|
||||
Specifies the particular size of blocks copied within the file when generating
|
||||
disk utilization. The value given may be followed by one of \fBb\fR (bytes),
|
||||
\fBkb\fR (kilobytes), \fBmb\fR (megabytes), or \fBgb\fR (gigabytes); if no unit
|
||||
is given, bytes are assumed. The default block size is 32768 bytes, or 32K.
|
||||
|
||||
.TP
|
||||
\-f \fIpath\fR, \-\-disk\-path \fIpath\fR
|
||||
|
||||
Use a file at \fIpath\fR as a file on which to generate disk activity. If
|
||||
\fIpath\fR does not exist, but its dirname (see \fBdirname(1)\fR) does, the
|
||||
file will be used as specified. If \fIpath\fR exists and is a directory,
|
||||
a secure temporary filename will be created within it and used.
|
||||
|
||||
.SH CPU UTILIZATION
|
||||
|
||||
If CPU utilization is enabled, a spinner process will be forked for each
|
||||
physical CPU detected on the system (or a fixed number of spinners, if the
|
||||
\fI-n\fR option is given.) All spinners will monitor \fB/proc/stat\fR's
|
||||
cumulative CPU utilization counters, attempting to keep total system CPU load
|
||||
at the desired level.
|
||||
|
||||
If other processes use a proportion of CPU time \fBless\fR than the chosen
|
||||
amount, lookbusy will use the remainder. If other processes use \fBmore\fR
|
||||
than the chosen amount, lookbusy will reduce its own usage to a minimum
|
||||
until the load drops below the chosen level.
|
||||
|
||||
The self-adjustment technique can take a relatively long time, by
|
||||
computational standards, to settle -- about 500ms on a 1GHz PowerPC CPU.
|
||||
Because of this latency, lookbusy should not be expected to defer to
|
||||
latency-sensitive applications quickly enough to avoid competing with them for
|
||||
CPU during periods of escalating consumption.
|
||||
|
||||
.SH EXAMPLES
|
||||
.TP
|
||||
\fBlookbusy \-c 10\fR
|
||||
|
||||
Attempt to use 10% of all available CPUs.
|
||||
|
||||
.TP
|
||||
\fBlookbusy \-c 75 \-n 2\fR
|
||||
|
||||
Attempt to keep up to two CPUs 75% utilized (if there is only one CPU, the two
|
||||
spinners will divide the load, though not necessarily evenly).
|
||||
|
||||
.TP
|
||||
\fBlookbusy \-\-cpu\-mode curve \-\-cpu-curve-peak 14h \-c 20\-80\fR
|
||||
|
||||
Maintain a CPU load of between 20% and 80%, on a curve repeating every 24
|
||||
hours (the default), with its peak at 2PM local time.
|
||||
|
||||
.TP
|
||||
\fBlookbusy \-c 20\-30 \-\-cpu\-mode curve \-\-cpu\-curve\-period 60m \-\-cpu\-curve\-peak 30m\fR
|
||||
|
||||
Maintain a CPU load on a utilization curve between 20% and 30%, reaching a
|
||||
peak every hour on the half hour.
|
||||
|
||||
.TP
|
||||
\fBlookbusy \-c 0 -m 512mb \-M 100\fR
|
||||
|
||||
Keep 512MB of memory occupied, with a memory stir iteration every ~100 usec;
|
||||
don't use any other CPU time.
|
||||
|
||||
.TP
|
||||
\fBlookbusy \-c 0 -d 2gb \-D 10 \-f /var/tmp/\fR
|
||||
|
||||
Generate disk traffic via a 2GB temporary file in /var/tmp/, with a 10
|
||||
microsecond pause between each block operation. Don't use CPU time.
|
||||
|
||||
.SH COPYRIGHT
|
||||
Copyright (c) 2006, Devin Carraway <lookbusy@devin.com>
|
||||
.br
|
||||
This is free software. You may redistribute copies of it under the terms of
|
||||
the GNU General Public License, version 2 or (at your option) any later
|
||||
version. See <http://www.gnu.org/licenses/gpl.html> for the text of this
|
||||
license. There is NO WARRANTY, to the extent permitted by law.
|
||||
|
||||
Reference in New Issue
Block a user