Latency monitoring is of critical importance for any website or application - an annoying user experience can be a death sentence for an otherwise enjoyable app.

This project explores an application of the t-digest data structure, introduced by Dunning and Ertl (repo here: https://github.com/tdunning/t-digest).

The t-digest data structure stores a compressed version of a collection of 1-D data. The structure allows for fast computation of percentiles and is mergeable, allowing for natural parallel computations on a platform like Spark.

Presentation slides can be found here: bit.ly/2sZoe3I