[smokeping-users] Smoke cut-off problem back in 2.4.0?

G.W. Haywood ged at jubileegroup.co.uk
Thu Jun 12 10:25:47 CEST 2008


Hi there,

On Thu, 12 Jun 2008, Tobias Oetiker wrote:

> > I've just upgraded to 2.4.0 from 2.3.5.  I never experienced the
> > issue with smoke getting cut off at the top of graphs (previous
> > thread), but now I have it.
>
> Earlier there was a bug that prevented smoke that was 'out of the
> picture' from being displayed ...
>
> So I am still trying to figure a good (automatic) solution here. At
> the moment I am using the median to scale the graph and just add a
> factor of 1.5 to the median to give some space to smoke. The idea
> is that I do not want the scale the graph into the bottom just
> because there is some high smoke ... but maybe this is wrong.
> ...
> anyone ?

I don't think there is 'right' and 'wrong' here.  These are very
difficult decisions to make for everyone else.  Networks, the ways
they are used, and their users all have individual characteristics
which will both differ greatly and change unpredictably with time.

My feeling is that some work is needed on the use of statistical
techniques, some of which have been applied to similar problems for
many decades.  The analyses could be made available as options.  The
focal point of this discussion could then move away from presentation
of information (and in many cases the best presentation will probably
become obvious), and towards the extraction of features which might at
present be hidden by non-random noise.

One thing which concerns me a little is the treatment of data samples
by RRDtool.  As I have noted in other threads the 'collapsing' of the
samples in order to give a fixed database size seems to me to result
in a tendency to mask details, and even in some cases to give a false
impression.  I'm not convinced that this is a good way to manipulate
the kinds of information with which we are dealing, and I would much
prefer that the data storage techniques and the data manipulation were
decoupled so that an investigation which is triggered by some sudden
revelation (such as a penny dropping) is not be hindered by the tools
themselves.

Having said all that I find that RRDtool does a good job for me now
that I have tinkered with it a little to suit my needs.  Thank you.

--

73,
Ged.



More information about the smokeping-users mailing list