Hi Greg!
----- "Greg Banks" <gnb@xxxxxxxxxxx> wrote:
> Greg Banks wrote:
> > G'day,
> Hmm, Thunderbird decided to Send instead of Paste.
>
> >
> > There was an article linked from Slashdot today
> >
> > http://queue.acm.org/detail.cfm?id=1809426
> >
> > which featured an interesting form of graph which the now-Oracle
> > Fishworks guys call a heat map. Basically, it's a plot with 3 axes:
>
> > x, y, and a z which is represented as the grey level of a pixel.
> > Their examples seem mostly to be x=time, y=latency or disk offset,
> and
> > z=number of samples, so that the (y,z) slice is basically a
> histogram
> > rotated up from the screen, and the chart as a whole represents how
>
> > that histogram changes over time.
> >
> > There's nothing new or magic about this. Audio engineers have been
>
> > using precisely this chart format for years, except that they call
> it
> > a spectrogram.
> >
> http://en.wikipedia.org/wiki/Spectrogram
>
> I was wondering if there was a feasible way to display such a chart in
> kmchart, assuming the existance of either a) a single counter metric
> with multiple instances representing histogram buckets, or b) multiple
> metrics representing histogram buckets.
We cannot do this *today* but this is the kind of thing I'm hoping to add
support for in the next big iteration of km^Wpmchart (kmchart just a shell
script nowadays, btw).
pmchart uses Qwt for its graphing. Qwt already supports spectrograms:
e.g. http://qwt.sourceforge.net/spectrogramscreenshots.html
... so, lucky us!, we even have code readily available to provide this.
The trickier parts of implementing it will be:
- extra UI work will be needed for creating/editing such charts - not just
the one metric per plot now, and specifying the colour map would need some
design thought & new code
- config file support for saving/loading views for these more "interesting"
chart types. This is well underway, it just got held up behind the cgroup
support in pmdalinux in the last few weeks (ongoing), but at least that
side of things will be solved soonish.
cheers.
--
Nathan
|