Reference to the mvar() function


mvar(<timeWindow>, <tsExpression>)

Returns the moving variance for each series over the specified time window.


timeWindow Amount of time in the moving time window. You can specify a time measurement based on the clock or calendar (1s, 1m, 1h, 1d, 1w), the window length (1vw) of the chart, or the bucket size (1bw) of the chart. Default is minutes if the unit is not specified.
tsExpression Expression that describes the time series you want moving variances for.


The mvar() function computes the moving variance of each time series over a shifting time window. For example, mvar(60m, ts(my.metric)) returns, at each point, the variance in the data values over the previous 60 minutes for each specified time series.

You can apply the sqrt() function to mvar() to get the moving standard deviation, for example: sqrt(mvar(120m, ts(my.metric))).


The following example computes the moving variance for the CPU usage user percentage over 5 hours.

mvar simple

We can see that the variance flattens out as we look at 5 days.

mvar 5d

See Also

Using Moving and Tumbling Windows to Highlight Trends

Detecting Anomalies With Functions and Statistical Functions