The graphing script is in Listing 2, 2396l2.orig.
A breakdown of the major parts of this script follows:

Line 12: trap command to remove working files if the user hits break

Lines 18 - 23: A function that will check if a file exists and 
has a file size greater than zero

Lines 35 - 52: Some variable definitions and help/usage information

Lines 60 - 90: Parsing the options, set switches or display 
information as requested. I find it helpful to include options for help 
(-h or -H), RCS log
information (-L) and version number
(-V) and debugging (-D); so I
use a standard template for most of my scripts with these basic
options, then add appropriate options for the job at hand.
For this script the additional options are: print without display
(-P), display the chart with ghostview, then 
offer the option of printing or not (-p).

Lines 92: Change the current directory to the one where the 
data files are collected.

Lines 93 - 105: Check for a passed file name; if no file 
name was passed default to constructing the name for today's data
file.

Lines 106 - 117: Validate the file name, making sure that the 
file exists, is readable and is non-zero in length.

Lines 121 - 161: Run the data file through an in-stream awk 
script, converting the timestamp to hours and hundredths so that the plot 
will have the data distributed evenly across the scale for each hour. If 
we left the minutes scaled to 60, the data points would be clustered
within the first .6 of the hour.

@bb:	If we find a value for user time that is lower than the prior 
observation, it is either a reboot or the counter rolled over. Since it is 
usually a reboot, I adjust the saved values to zero to reflect the
counter reset and set the reboot data point to 100 to generate a reboot
indication on the chart. On normal observations the reboot data point is
set to -1, so the plotted point will fall outside of the chart and not
appear.

@bb:	Calculate the gain for each counter and divide by the number of seconds
since the last records and write the values out to a temporary file with
the converted timestamp. Save the value of each variable for use with the
next record.

Line 164: Get the name of the host. I'm assuming 
that this is the machine that the data was generated on.
This also assumes that the data being charted is for the machine it is
being charted on. If I were to go with a central data collection
and analysis system, I would embed the originating host name in the
file name.

Line 165: Clip the date out of the file name of the input for 
use in chart titles.

Lines 166: Display a progress message, showing the system 
and date. 

Lines 170 - 179: Define the gnuplot script line that sets the 
output device. If the user requested hardcopy, set the "set 
terminal" and "set output" for PostScript output; 
otherwise, set the terminal as X11 for display in an X11 window. We also
"set label" for the Y axis for hardcopy and leave it null for 
X11, as it gets displayed incorrectly on the display. The X11 display gets a 
pause of ten seconds to allow some time to view the chart. Without the 
pause the graphic display is cleared too quickly to be seen.

Line 183: Start echoing the in-stream script as stdin to the 
gnuplot program to generate the chart.

Line 184: Set the previously defined "set
terminal" command.

Lines 185 - 187: Turn on plotting with lines, turn off the 
grid overlay, and turn on the key legend display for labeling the lines.

Line 188: Set the chart title, using the system name and the 
date from the input file name.

Lines 189 - 191: Set up the X-axis variables: label, range 
of values and range of tic marks.The x scale is used for the time of day and 
is labeled as such, from 0 to 24 with tic marks for every hour.

Lines 192 - 194: Set up the Y-axis variables: label, range of values 
and range of tic marks.The Y scale is for the percentages and as such runs
from 0 to 100, but to allow room for the legends in the upper right
corner the scale is set to 0 to 140, with tic marks every 10 points
from 0 to 100.

Lines 195 - 200: Define the plots, input file, columns to 
use, title for the legend and plot type. The reboot plot is done with points 
rather than a line, as it refers to events rather than trends (we hope).

Lines 201 - 202: Pause, if appropriate, and quit.

Line 203: The EOF marks the end of the input stream.

Lines 204 -217: If hardcopy and no display 
before printing is requested, go right to lpr. If a display before 
printing is requested, send the PostScript file to Ghostview, then prompt 
for whether to print or not. Clean up the PostScript file when done.

Line 218 - 221: I have found that I want to look at the work 
files when I'm debugging, so If I set the debug switch
(-D), keep the data file; otherwise, trash it.

@b:The other scripts work on the same basic pattern with some additional
twists for the scripts for disks and interrupts, as I
try not to generate graphs for non-existent disks or unused interrupts.
I've pulled the descriptive text for the used interrupts out of the
/proc/interrupts file to add a little more detail to the charts.