Overview

Edit online

OMEGAMON® AI Insights can detect and visualize performance anomalies within z/OS components and applications. It draws its intelligence from sophisticated machine learning algorithms that analyze data collected by OMEGAMON monitoring agents.

Note: OMEGAMON AI Insights is not available separately. You must have one of the following product offerings that contains it:
  • IBM Z® OMEGAMON AI for JVM 6.1
  • IBM Z OMEGAMON AI for Networks 6.1
  • IBM Z OMEGAMON AI for z/OS 6.1
OMEGAMON AI Insights machine learning process consists of two steps:
  1. Train: Historical data is used to build or update forecast models. This step does not occur in real-time but should be done regularly for getting seasonality trends. The Training process needs some data over a reasonable period to be able to build a forecast model. The standard minimum period of training is fourteen days. OMEGAMON AI Insights models are generated immediately, however it is only after the standard period of training that these forecasts should be considered.
  2. Predict: This process uses the models to determine if the current metric is beyond the forecast. It creates an anomaly and calculates the severity level. This step is generally scheduled for every hour.

OMEGAMON monitors the behavior of applications and components by tracking various performance metrics referred to as attributes. In the process, it collects enormous amounts of data. With this data, OMEGAMON AI Insights can arrive at models that predict the seasonality of various key performance indicators across different types of workloads.

The generated models start providing predictions as soon as they are installed, however, the model matures over time through a training process that uses historical data.

Once a model has matured, you can begin comparing the forecasts it generates with live key performance indicators. For instance, you would know if the CPU utilization measured on a given day is abnormal if it exceeds the value predicted for that day by the model.

All data used by a model is stored and manipulated in Elasticsearch, which in turn sources it from OMEGAMON Data Provider.

The live data, along with the predicted seasonality patterns output by a model, are presented through visually appealing dashboards on Elastic Kibana. Through these dashboards, you can easily detect abnormal behavior, and promptly troubleshoot by drilling down into specifics.

In addition to providing visual warnings, OMEGAMON AI Insights can also email alerts to stakeholders based on the severity of an anomaly.