The server aggregates results, except when the query indicates not to, by using a pivot. The mechanism for aggregating the data depends on the type of metric. A pivot directs the query processor to not aggregate one component of the query data. A pivot can be based on time, server instance, a specific dimension, or histogram ranges.

  • If no pivot is specified, the query returns a single number that represents the aggregation of all matched data. For example, a query with no pivot might return the total number of operations that have completed today.
  • A single pivot results in one-dimensional data, such as a time-based chart with a single line or a simple bar chart.
  • Two pivots results in two-dimensional data, such as a time-based chart with a separate line for each server instance, or a stacked bar chart that shows the number of completed operations broken down by server and operation type.
  • Three pivots results in three-dimensional data, such as a stacked, grouped bar chart that shows completed operations broken down by server, operation type, and result.

Beyond aggregating multiple samples into one, the data returned by a metric query can be further manipulated.

You can scale queries on the count statistic to return the count of events per second, per minute, or per hour. Counts of histogram values can be returned by a percentage of the total. For instance, instead of returning the raw count of operations that took longer than 50 milliseconds to complete, the results could return as the percentage of all operations that took longer than 50 milliseconds to complete. A value of 0.02% is more meaningful than a value of 40