dist-avg

Input Stack:

query: Query

⇨

Output Stack:

TimeSeriesExpr

Compute the average recorded value for timer and distribution summary metrics by dividing the total recorded amount by the count of recorded measurements. This provides the true average per individual measurement, not per time series or instance.

Parameters¶

query: Query for timer or distribution summary metrics

How It Works¶

This operator automatically combines the appropriate statistics to calculate per-measurement averages:

For Timers:

# Manual calculation:
name,http.req.latency,:eq,statistic,totalTime,:eq,:and,:sum,
name,http.req.latency,:eq,statistic,count,:eq,:and,:sum,
:div

# Equivalent using :dist-avg:
name,http.req.latency,:eq,:dist-avg

For Distribution Summaries:

# Manual calculation:
name,request.size,:eq,statistic,totalAmount,:eq,:and,:sum,
name,request.size,:eq,statistic,count,:eq,:and,:sum,
:div

# Equivalent using :dist-avg:
name,request.size,:eq,:dist-avg

Metric Requirements¶

The metrics must be instrumented using Timer or Distribution Summary classes that collect both total and count statistics. These classes automatically record:

totalTime/totalAmount: Cumulative sum of all measurements
count: Number of individual measurements recorded

Examples¶

Computing average request latency per request:

Before	After

name,playback.startLatency,:eq	name,playback.startLatency,:eq, :dist-avg

Simple usage pattern:

nf.cluster,foo,:eq,
name,http.req.latency,:eq,
:and,
:dist-avg

Computing average by auto-scaling group:

nf.cluster,foo,:eq,
name,http.req.latency,:eq,
:and,
:dist-avg,
(,nf.asg,),:by

Comparison with Other Averages¶

Operator	Denominator	Use Case
`:avg`	Number of reporting time series	General time series averaging
`:node-avg`	Number of unique instances	Per-instance average
`:eureka-avg`	Number of UP instances in Eureka	Per-active-instance average
`:dist-avg`	Number of recorded measurements	Per-measurement average

:avg - Standard average across time series
:node-avg - Average using total instance count
:eureka-avg - Average using UP instances in Eureka service discovery