Smoothed curves look very eye friendly! One thinks to understand the results much easier which is probably true for monitoring but very dangerous for analysis. There is a big danger to overlook or misinterpret real issues.


In the example below you can see why:

  • This is two time the same analysis. The upper one is original the lower one is smoothed.
  • Let's just take one Latency Peak (red) on the far left in the trace

    In the original representation the peak happened at 1:47:22 and is 4.94 ms high.
    This is a relevant peak which could have caused troubles

    In the smoothed version the same peak no longer falls on.
    If one determines his time, he was held only at 3:30 and is only 2:32 ms high.
    So looks to be  not a relevant issue and if so, it happened 2 hours later!



Why are the results of smoothed curves so wrong?

Smoothing works on moving averages and enlarged intervals. These two mechanisms eliminate peaks and normalize the measured values which thereby become smaller. Such information is probably capable in management presentations but do not have to be used in analysis work.

BVQ only switches to smoothing on long time periods of 1 week or longer. This is done to decrease the number of measurement points shown which could confuse on so large time periods.

It is easy to switch on or of smoothing in BVQ. Just use the option panel to set values for fixed interval and moving average