dist-stddev

Input Stack:

query: Query

⇨

Output Stack:

TimeSeriesExpr

Compute the standard deviation for timer and distribution summary metrics based on the recorded samples. This calculates the standard deviation of the individual measurements that were recorded, showing the variability of the actual sample values.

Parameters¶

query: Query for timer or distribution summary metrics

How It Works¶

This operator automatically combines the appropriate distribution statistics to calculate per-measurement standard deviation using the mathematical formula:

σ = √((totalOfSquares × count - totalTime²) / count²)

For Timers: - Uses totalTime, totalOfSquares, and count statistics - Calculates standard deviation of individual request durations

For Distribution Summaries: - Uses totalAmount, totalOfSquares, and count statistics - Calculates standard deviation of individual measurement values

Metric Requirements¶

The metrics must be instrumented using Timer or Distribution Summary classes.

Examples¶

Computing standard deviation of request latency measurements:

Before	After

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

Simple usage pattern:

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

Manual Equivalent¶

The :dist-stddev operator simplifies this complex manual calculation:

statistic,count,:eq,
:sum,
statistic,totalOfSquares,:eq,
:sum,
:mul,
statistic,totalTime,:eq,
:sum,
:dup,
:mul,
:sub,
statistic,count,:eq,
:sum,
:dup,
:mul,
:div,
:sqrt,
nf.cluster,foo,:eq,
name,http.req.latency,:eq,
:and,
:cq

Comparison with Other Standard Deviations¶

Operator	Measures	Use Case
`:stddev`	Variation across time series	Group consistency analysis
`:dist-stddev`	Variation within measurements	Distribution spread analysis

:stddev - Standard deviation across time series (different calculation)
:dist-avg - Average for timer/distribution metrics
:dist-max - Maximum for timer/distribution metrics
:sum - Total values across time series
:by - Group results by dimensions