[time-nuts] Thunderbolt Data Plotting
hmurray at megapathdsl.net
Fri Jul 11 14:41:39 EDT 2008
> Well - I have lots and lots of raw numbers, and no really good way of
> plottinng it. I'm hardly a linux guru, but would gnuplot be a suitable
> tool for graphing data like the snipper below?
It's what I use.
gnuplot has zillions of options. You will probably have to read (or at least
skim) the whoie manual to get a feel for how to do things. Once you know
what you are looking for, the online help works pretty well.
An important question is how to organize your raw data. I copied the setup
from ntpd's "day" mode. It uses 1 file per day with filenames like
loopstats.20080628. The first two columns of the file are the modified
julian day and seconds within the current day. For example:
54658 64083.850 -0.000002000 118.700 0.000016330 0.009421 9
54658 64149.858 -0.000022000 118.700 0.000016849 0.008813 9
54658 64215.865 -0.000040000 118.699 0.000016956 0.008251 9
54658 64280.873 -0.000049000 118.698 0.000016187 0.007726 9
54658 64346.881 -0.000064000 118.697 0.000015979 0.007236 9
You can also link loopstats (no date) to the latest one so tail loopstats
gets you the latest/current one without bothering to figure out what the date
I generally set things up so I can easily plot the last two files. That
gives somewhere between 24 and 48 hours so I always have some context if an
interesting quirk happens at the start of a day.
I have a handful of scripts that grab log files from several machines and
setup links to the latest two. For example, foo-loop goes to
foo/loopstats.20080711 and foo-loop-1 goes to foo/loopstats.20080710 Then
the gnuplot include files just refer to foo-loop and foo-oop-1 When making
graphs, I ignore the first column and add a 24 hour offset on one file to the
That generally works OK for looking at events with a time scale of hours. I
haven't tried looking for long term (days, months) trends. If/when I want to
do that, I'll probably do some pre-processing and make a new file with one
line per day or hour.
These are my opinions, not necessarily my employer's. I hate spam.
More information about the time-nuts