Distributions are a metric type that aggregate values sent from multiple hosts during a flush interval to measure statistical distributions across your entire infrastructure.
Global distributions are designed to instrument logical objects, like services, independently from the underlying hosts. Unlike histograms which aggregate on the Agent-side, global distributions send all raw data collected during the flush interval and the aggregation occurs server-side. Because the underlying data structure has not been aggregated and represents raw data, distributions provide two major features:
Calculation of percentile aggregations: Percentile aggregations (p50, p75, p90, p95, p99) are calculated from the raw data across all hosts, and are therefore globally accurate.
Customization of tagging: This functionality allows you to control the tagging scheme for metrics for which host-level granularity is not necessary (e.g. transactions per second for a checkout service).
See the Developer Tools section for more implementation details. Please note that since distributions are a new metric type, they should be instrumented under new metric names during submission to Datadog.
Like other metric types, such as
histograms, distributions have the following aggregations available:
avg. Distributions are initially tagged the same way as other metrics (with custom tags set in code) and are resolved to any host tag based on the host that reported the metric. You can also calculate percentile aggregations for a set of tags (up to ten) specified on the Distribution Metrics page. This provides aggregations for
After electing to apply percentile aggregations on a distribution metric, these aggregations are automatically available in the graphing UI:
Distributions provide functionality that allows you to control the tagging for metrics where host-level granularity does not make sense.
To customize tagging, hover over your metric in the table, and click on the pencil icon to edit. In the modal that pops up, select Custom…. There is a whitelist of the tags you have defined in code by default. You can remove any of these tags or add any host-level tags back in.
Note: The exclusion of tags is not supported in the whitelist-based customization of tags. Adding tags starting with
! is not accepted.
Distribution metrics with percentile aggregations (
p99) generate custom metrics or timeseries differently than gauges, counts, histograms, and distributions with nonpercentile aggregations (
avg). Because percentiles aren’t reaggregatable, Datadog preserves five timeseries for every potentially queryable tag combination. This is different from the number of custom metrics generated from gauges, counts, histograms, or distributions with nonpercentile aggregations (dependent on the unique number of tag value combinations that appear in your data).
For more information on counting custom metrics created from gauge, count, histogram, or distribution metrics with nonpercentile aggregations, refer to the Custom Metrics page.
Documentation, liens et articles supplémentaires utiles: