Potato On Mars is a daemon created to monitor server load and detect anomalies on the time series by compraing regular patterns recorded and new patterns.
Potato On Mars simply record a lot of data to create its data base for regular patterns.
Regular patterns are discretized in 24-letters-long words with one letter for each hourwith the HotSax algorithm
Then when a new data comes out the software makes a new 24h-long pattern with 23 other data points, discretizes it and simply computes an euclidian distance between regular patterns and the new pattern.
The distance is then compared to a threshold. If it is greater then an alert is raised.
We used different mathematical methods related to time series in our analysis phase.