What is Morgoth?
Morgoth is open source software that provides simple and flexible anomaly detection for time series metric data.
Getting Started
Follow a tutorial on using Morgoth here
Why use Morgoth?
- Morgoth is simple
- Morgoth integrates with your existing infrastructure
- Morgoth has powerful anomaly detection built-in
- Morgoth is extensible
Powerful anomaly detection built-in huh?
Have a look at Detection Framework for a detailed walk through of how the anomaly detection works.
One powerful component of Morgoth is its ability to use many different algorithms. Using standard deviations or 3-sigma algorithms is a very common way of detecting anmalies in metric data. These techniques assume the data they operate on follows a Gaussian distribution. Unfortunately much metric data is not Gaussian. Take cpu usage for example. Servers perfoming work from a queue tend to be nearly idle then spike to 100% cpu usage and then drop back down. Most of the time the server is either near 0% utilization or 100% following a bimodal distribution. If cpu usage were gaussian then the cpu would spend most of the time around 50% utilized and rarley 10% or 90% utilized.
Several algorithms that ship with Morgoth are non-parametric, aka they do not assume a normal distribution of the data. These alorithms often perform better on certain data sets. Morgoth lets you choose the alorithms and have them all vote on whether a window of data is anomalous.
Join the community
Mailing listIRC on freenode in #morgoth