Within Datadog, a graph can only contain a set number of points and, as the timeframe over which a metric is viewed increases, aggregation between points occurs so that the number of points remains under that set number. Thus, granularity is lost as the timeframe increases. For instance, for a four hour time window, data is aggregated to have one value per minute for a line graph, and one value per two minutes for a bar graph. As you “zoom out” (i.e. select a larger timeframe) the data shown on the graph represents a longer time period.
When bars are displayed the rollup interval is more obvious:
You can manually append the
.rollup() function to your query to adjust the method and granularity of time aggregation. Datadog rolls up data points automatically by default, averaging values in the rollup interval for
COUNT metric types.
Note: If you query your metrics through the UI of a Datadog widget, an in-application metric types modifier is added automatically to your
COUNT metric types. This changes the
.rollup() behaviour: values are summed without any interpolation.