01-28-2013 07:48 AM
Analytics detects, in real-time, metrics that have activity outside their baseline, and labels this activity as “abnormal behavior”.
When Analytics identifies abnormal behavior it applies several calculations to alert you of real problems and avoid false alarms. Analytics is able to automatically filter out false alarms by using several advanced algorithms, including the following:
- Auto-correlation of business and infrastructure. For example, if there is abnormal CPU usage, but all the relevant business transactions are behaving normally and are not affected, Analytics will not send an alert.
- Identify noisy metrics. Noisy metrics are less trustworthy as they will breach their baseline more often than other metrics. Analytics can learn from past experience how noisy a metric is; the more noisy the metric, the greater the chance that the metric does not represent a real problem.
- Learn from historical data. The more monitoring you have, the less noise you get. The more that you monitor your application, the more data Analytics will have to calculate the real status of the application, and will produce more accurate results. Analytics will be better at detecting false alarms in the system and will be able to alert you only on the issues that matter to you.