AIMS analyzes all performance parameters in close to real-time and generate normal behavior patterns for all of these with different time-resolutions. A normal behavior pattern means dynamic thresholds which are unique for every performance parameter that AIMS is monitoring. As seen in the screenshot above, the normal behavior pattern is the blue area that represents a dynamic upper and lower threshold.

AIMS starts by learning patterns for up to 48 hours and can then start comparing all hours of the day between two different days. Then AIMS continues to learn daily patterns while building weekly patterns, and to be able to compare weekly cycles AIMS need two full weeks of data. This is why the anomaly detection is enabled 14 days after connecting the agents, as these normal behavior patterns (dynamic thresholds) are an important component of the anomaly detection.  After this AIMS continues to build patterns for months and finally years.

All organizations have some kind of business cycles -> daily, weekly, monthly and / or yearly. AIMS is able to learn these different cycles and can then easily spot deviations from normal behavior on the different performance parameters.

Looking at the example above, this is how the CPU time on a database is supposed to behave on a Monday between 12:00 AM and 23:59 PM, in the first week of the month. Next Monday in the next week might look completely different, as the organization might have a monthly cycle that impacts the different weeks of the month.

In the screenshot, the red line represents the actual measured values. As long as these are within the normal behavior pattern, your system / parameter is behaving as expected. Then at 12 PM (screenshot) there is a deviation from the normal behavior pattern, that is then classified by AIMS using multiple algorithms. If the trend shows that this is a deviation that will continue to grow, AIMS will correlate this behavior towards all other parameters to check for potential impact. If impact is found, an anomaly alert is sent to the user.

AIMS can at any time be put in re-learning mode if there are any changes to the data that is being monitored. Read more about this here.