Automatic z/OS Performance Analytics
Advantages of ConicIT for mainframe during the first few weeks
During the first few weeks of ConicIT installation, the system is not yet capable of producing proactive alerts (anomaly alerts) based on statistical models and algorithms, simply because there’s not enough history to base the “normal-behavior” model upon.
This might bring us to the false conclusion that ConicIT benefits starts only 3 weeks post installation. After all the proactive alerts that are based on unique statistical algorithms is very powerful.
However, ConicIT actually does have important advantages that it provides from the very first moment after installation and configuration:
- ConicIT can aggregate data from different sources and present them in a combined way.
- ConicIT can produce calculated variables which are hard to calculate manually and can provide important information.
- ConicIT can send static-alerts, which are not based on statistics. So these alerts are not based on the ability of ConicIT to study the normal behavior, but they can still provide important alerts, and provide helpful information for analyzing the alerts.
We’ll describe some of these advantages within this document.
First Fault Performance Problem Resolution
ConicIT was founded based on the observation of a fundamental weakness of commonly used performance monitoring systems; too much data is available and there are no accurate analysis tools. We have reached a point where there is too much data for a human to process. This data overload is caused by two factors: the first being the time lag between problem occurrence (when meaningful analyzable factors are available) and problem detection when the data is only symptomatic. The second is that using static thresholds as an alerting mechanism leads to inaccurate alerts and mistrust of the alerting system. It became clear that if it was possible to predict problem occurrence before the data becomes dominated by symptoms, we could mitigate both problems – provide accurate alerts and pinpointed problem data.