[rrd-users] trying to understand the relationship between source data, what's in rrd and what gets plotted

Alex van den Bogaerdt alex at ergens.op.het.net
Sun Jul 22 11:28:31 CEST 2007

On Sat, Jul 21, 2007 at 02:52:55PM -0400, Mark Seger wrote:

> >> Just put the dots (or lines) where they belong and if some fall on top 
> >> of each other and you only see 6, that's just fine.  Where I'm coming 
> >> from, and what I always do with performance related data, is look at a 
> >> broad range of time, perhaps a day wide.  I want to look at everything 
> >> from cpu load, network, disk, memory usage, infiniband loading and a 
> >> whole lot more.  I want to be able to see at a glance if there are any 
> >> spikes where there shouldn't be or drops that also shouldn't be there.  

Perhaps a day wide: 86400 seconds.

See at a glance if there are any spikes or drops -> you want to
see extremes.

Now create a graph of 432 pixels wide. In each column, display
the minimum and maximum of 200 rates, as a VERTICAL line.  If
this line starts high enough (lowest rate high enough) and stops
low enough (highest rate low enough), you can see, at a glance,
you do not have to examine these 200 rates.

If the line starts too low or ends too high, you need to zoom in.

You will have 432 of such vertical lines, representing 86400 rates.

The vertical line is in reality a stack of areas: First an invisible
area to get an offset from the X-axis to the minimum rate, then a
visible area representing (maximum-minimum) thus displaying both of
these values.

Next picture: again 432 pixels wide, now showing 432 rates. No need
to look at minimum and maximum, as min(one rate) equals max(one rate)
equals average(one rate).

The resulting line will go lower than usual (a rate was too low) or
higher than usual (a rate was too high).  This is what you were looking

Now, please tell me what this example is missing.

Alex van den Bogaerdt

More information about the rrd-users mailing list