Hi Niko

> The old behaviour was cutting both the higher values *and* the smoke
> leading up to them.  This was a bug that got fixed in 2.3.6 so that the
> smoke is now visible.

To me it looks like the y-scale has a smaller range in the new
version. But maybe the changed smoke representation fools me.
I'll try to make a few snapshots when I upgrade my main server
(before and after the upgrade) to compare them. Unfortunately
this probably will not happen before the first upgrade release
of lenny.

> The graph intentionally scales with the median values, not the peaks,
> and I believe the Etch version did that too. This may become more
> configurable in the future; currently I think the only related options
> are 'unison_tolerance' and 'logarithmic'.

I agree to cut off the most highest peaks, maybe the top 5 or
even 10%. But if you look at the graph I attached to my first
mail it cuts off nearly all of the values which makes the graph
lose a lot of information about the time distribution (you can't
see an upper envelope of the values anymore).

Increasing the unison tolerance doesn't help here. Actually the
submitted graph already was created with unison_tolerance=3.
I also have tried a value of 5 now. And a logarithmic scale does
not fit the nature of network delays for me.

Regards

Uwe




-- 
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]

Reply via email to