Build resilient systems at scale
28–30 October 2015 • Amsterdam, The Netherlands

The definition of normal - an intro and guide to anomaly detection

Alois Reitbauer (Dynatrace)
14:40–15:20 Thursday, 29/10/2015
Location: Emerald Room
Average rating: ***..
(3.93, 27 ratings)

Prerequisite Knowledge

basic understanding of statistics and some experience working with operations performance data.


The session looks into how anomaly detection works by discussing these key concepts:

  • What is normal behaviour?
  • How are expectations about future behaviour derived from data?
  • How do anomaly detection algorithms work including trending and seasonality?
  • How do these algorithms know whether something is an anomaly?
  • Which algorithms can be used for which type of data?

The talk will work with simple-to-understand hands-on examples. I will cover basic concepts like distribution models, exponential smoothing, hypothesis testing etc. I will also provide clear guidance on which types of algorithms to use for failure rate analysis, response times, load, and other key operational metrics.

Photo of Alois Reitbauer

Alois Reitbauer


Alois Reitbauer is chief evangelist at Dynatrace. Alois is fanatic about monitoring, DevOps, and application performance and has spent most of his professional career building monitoring tools and speeding up applications. He is a regular conference speaker, blogger, book author, and sushi maniac.