Response times in raw data do not match graphs

First it should be noted that the 'meta data' columns in the raw CSV export are cumulative times at each stage of the operation. The numbers shown on the graphs, however, are discrete times for each.

Secondly, the graphs are rendered from rolling averages calculated from the raw data, instead of the raw data itself.

For example: A sensor monitored at a 1 minute interval generates 1440 data points per day. In ~600 pixel wide graph we obviously cannot include all of those data points.

Instead we use rolling averages of the sample data to produce the graphs. Rolling averages will naturally smooth out peaks and troughs in the graph data. You can see the most extreme example of this by viewing a 12-month graph - we render those using 24-hour rolling averages.

Also note that the numbers shown in the graph legend are 'last' values - they are the last calculated rolling average for that particular sensor. They are not an average for the entire graph period.

For these reasons, you will see differences in between the timing data collected by the sensors in the raw data export, and the numbers shown in the performance graphs.

Not what you were looking for? Try a search:

Ninja Tip: trace* will match traceroute.

Also in this topic: