Sum

a sum metric represents a cumulative value that changes over time by either only increasing (monotonic) or both increasing and decreasing (non monotonic) a sum metric can be of delta or cumulative temporality temporal aggregation (aggregation within the same time series) when you specify an interval (e g , 300 seconds), raw data points from the same time series are grouped into buckets based on that interval within each bucket, an aggregation function is applied to the points that fall inside the defined time window for sum metrics, the most meaningful aggregation operations are 	∙ rate per second rate of change 	∙ increase absolute increase during the interval 	∙ sum sum of all points when the metric temporality is delta in this case, each point already represents a change, so summing all delta values inside the bucket gives the total increase for that interval how rate and increase are calculated for each interval time window, the system evaluates data points differently based on the sum temporality for cumulative sum, the system selects the maximum (typically the last) value inside the time window this represents the reported counter value within that window and is compared with the value from the previous time window for delta sum, each point already represents a change, so the system performs a sum of all values inside the time window that total represents the increase for that window, and no comparison with the previous window is required the system automatically detects the metric temporality (cumulative or delta) and applies the appropriate aggregation formula internally increase cumulative increase = current window value − previous window value delta increase = sum(all values in the current time window) rate cumulative rate = (current window value − previous window value) / (timestamp current − timestamp previous) delta rate = sum(all values in current time window) / time interval (seconds) where current window value → latest value in the current time window (cumulative) previous window value → latest value in the previous time window (cumulative) time difference → measured in seconds between two windows note for counter reset handling in cumulative metrics, if the difference between two time windows is negative, the system treats it as a reset and sets the computed value to 0 spatial aggregation (aggregation across multiple time series) spatial aggregation combines values from multiple time series at the same timestamp or interval bucket it is applied after temporal aggregation is completed (e g , rate or increase per series) and aggregates across all matching series if you select sum for spatial aggregation without any grouping, the system aggregates all available series into a single series