Welcome to my collect page
I no longer maintain collect -- in fact no one is maintaining collect, or, for that matter, Tru64 any more. Tru64 is dead, haven't you heard?
December, 2003
Bernard Laforgue has made a kit
available to install the old Perl-based
collgui version on Tru64 UNIX V5.1B. There is also
a readme for this kit.
Version 2.0 Information
The V2.0 kit has a few problems, some of which I have fixed.
Go to my Collect V2.0 Page for details.
Introduction
collect is an tool to collect operating
system and process data under Digital UNIX, versions 3, 4, and 5.
It is intended
- to be as cheap as possible (use few resources)
- to gather as much performance data as reasonably possible
- to be highly flexible
collect gathers data for 'subsystems', which are
|
|
lsm
network
cpu
|
|
tty
filesystems
message queue
|
data can be collected for any combination of subsystems.
collect has two modes: collection and playback. In collection mode,
it gathers operating system and process data, and writes it to standard
output, to a binary file with a special format, or to both.
In playback mode, collect reads from a previously written binary
file and writes to standard output. The format of the data when written
to standard output is identical, whether in collection or playback mode.
Look here for an example of collect output.
Digital UNIX versions
Collect runs on version 3, 4, and 5 (steel) of Digital UNIX. Because
Digital UNIX has been evolving, some statistics, specifically for disks, have
first become available in V4 or even V5. You should not expect to all fields
to be meaningful for V3 or V4. In particular, in V3 only the fields
R/S, RKB/S, W/S, and WKB/S are valid. In V4 AVS, ACTQ, and %BSY contain
data, but they are known not to be reliable.
The fields AVW, WTQ, and %WT are zero.
A patch will soon (Nov, '98) be available
for V4 which will finally make the disk statistics reliable.
Currently, only in V5 are all fields valid.
Associated Tools
Accompanying collect, are two highly-integrated tools,
collgui,
a graphical front-end, and cfilt, which allows almost completely
arbitrary
extraction of data from the output of collect to standard output.
cfilt is not just a tool for collgui to use, though. It
formats the requested values in a one-line-per-sample format, space delimited,
which can be imported, for example, into an EXCEL spreadsheet effortlessly.
See the cfilt manual page for
more information.
collgui is a labor-saving tool which allows a user to quickly
analyze collect data. Behind the scenes, it spawns collect
in playback mode (you must have a binary collect file) to read the collect data. This is filtered by cfilt. Finally, the output from
cfilt is fed to gnuplot, a public-domain plotting utility,
which can display graphs in a variety of ways, including X11, PostScript,
and PPM (portable Pixmap).
Perl
Because collgui and cfilt are both perl scripts, and
collgui is additionally dependent on the perl Tk extension, if you
want to use these tools, you must have Perl, and additionally,
you must have the Perl/Tk extension to run collgui.
If you don't have Perl and Perl/Tk, I strongly
recommend you get and
install one of the setld perl/tk
kits
that I have made available. This is a complete perl kit and has no
dependencies on collect. Therefore you can use it completely
independently of collect.
The following topics are available for more information:
Rob Urban
<urban (at) unix-beratung (dot) de>