Fine-grained access to summary statistics — Librato Blog

Fine-grained access to summary statistics

We're happy to announce that as of today we're providing our users with much finer-grained controls over what summary statistics (e.g. min, max, average, etc) are used when creating a correlation/instrument in Librato.

Our users have always had the option of submitting multiple summary statistics for a given metric as a single measurement using the form <min, max, sum, count>. This is useful for metrics that are sampled at a higher frequency than reported to Librato e.g. timing the latency of each request made to a web service. We also use this format when generating our historical aggregations at coarser resolutions (i.e. 1m, 15m, 60m).

Previously however, access to most of these summary statistics have only been available through our API. If you were visualizing your metrics through our UI you were restricted to toggling between the sum or average of all samples, and this decision was a global setting on a metric that could not be varied from correlation to correlation. Until now:


Now when adding a metric to an instrument you can specify which summary statistic to use (defaults to the global metric setting), and in the case of a wildcard source additionally choose a second summary statistic to calculate across the sources.

API Support

We build our application on top of the same public APIs available to our users, and while implementing this feature we added a couple of new parameters that may be of use when querying measurements from our API. You can more find information about the group_by and breakout_sources parameters in our API documentation.

SDK Support

We've also released an update (v0.3.0) to our JavaScript SDK for embedding charts in 3rd-party applications. This new version correctly renders charts configured with the new summary statistics.